About the Correlation Between Crude Oil Corrosiveness and Results From Corrosion Monitoring in an Oil Refinery

During the last decades, cost pressure on oil refineries has grown continuously, particularly in Europe, where consumption of crude oil products is decreasing more and more.¹  As a consequence, oil refineries are looking for ways to increase their margin, and reducing crude oil costs is an important key factor. These costs can account for about 80% of the total refinery expenditures.²  One possible way to cope with this challenge is to purchase cheaper crudes on the global markets, such as crudes with elevated sulfur content and/or acidity. These “opportunity crudes” are, however, often difficult to process owing to their increased tendency for corrosion, fouling, coking, and unfavorable boiling point ranges. Furthermore, one corrosion issue per week worldwide occurs statistically that leads to a severe incident such as sudden leakages, e.g., resulting from pipe ruptures.³  These facts emphasize the need for corrosion control in petrochemistry, where corrosion monitoring is one important approach to maximize equipment integrity. Other than traditional techniques such as, e.g., intrusive weight loss coupons, new wireless sensors were developed in the last years. They monitor pipe or vessel wall thickness and, thus, corrosion online by ultrasonic testing (UT). There is, however, little experience with this new monitoring approach. Open literature provides few articles about the general sensor technique, its advantages, and typical measuring data.^4-8  This study aims to discuss not only advantages and challenges of this new technique by focusing on measuring data but also its correlation to empirical data and crude run history of an oil refinery. One important question is, for instance, if measured corrosion can be linked to crude oil parameters that represent the corrosion potential such as total sulfur content or total acid number (TAN). If the UT sensor data can be used to predict corrosion for a certain crude oil diet, this would be an important step forward to reduce inspection and maintenance costs. Also, opportunity crudes could be processed by a safer and more economic manner.

The first units of each oil refinery are crude distillation unit (CDU) and high vacuum unit (HVU) where crude oil is distilled in two steps (at atmospheric pressure and at vacuum) into its major fractions. There are a variety of corrosion mechanisms that play a role in these units and that usually have a very different appearance at different locations. This study focuses on the following main high-temperature corrosion mechanisms in CDU and HVU hydrocarbon streams.

Sulfidation or sulfidic corrosion is the reaction of steel and other alloys with reactive sulfur compounds such as H₂S from the fluid (e.g., crude oil) in high-temperature environments.⁹  Being observed at temperatures above approximately 260°C, sulfidation corrosion rates increase with temperature, peaking at approximately 450°C. Both carbon steel/low-alloyed steel and stainless steel can be affected by this mechanism, which usually results in general wall thickness reduction. The modified McConomy curves predict the influence of temperature, disclosing that increasing Cr and Ni alloying contents in the bulk material can reduce corrosion rates significantly.^10-11  This relation was extended to hydrogen/H₂S-containing environments by the modified Couper-Gorman curves.^11-12 

Dependent upon bulk material, flow conditions, and chemical composition of the fluid, the reaction product iron sulfide (FeS) can form a stable, protective sulfide scale eventually reducing the corrosion rate. But this passive layer can be damaged or even removed completely, which is influenced by flow conditions, temperature, content/type of sulfur compounds and naphthenic acids, and hydrogen/H₂S content, etc. Moreover, a further influencing factor is the flow regime (e.g., two-phase and/or turbulent flow) that can cause localized sulfidation.

In addition to H₂S, mercaptans (or “thiols”) are considered to be some of the most reactive and, hence, most corrosive sulfur species. Their concentration depends strongly upon crude oil and boiling point range. Experiments showed that mercaptans are most reactive somewhere between 235°C and 300°C where they can accelerate sulfidation. Thus, particularly middle distillate loops, such as light and heavy gas oil, are susceptible to mercaptan corrosion. A minimum mercaptan concentration of 100 ppm to 200 ppm was found to be necessary to enable mercaptan corrosion.¹³  Furthermore, it was suggested that corrosion rates depend upon mercaptan type and that mercaptans start to decompose at temperatures between 300°C and 400°C.^13-14 

Each crude oil contains naphthenic acids, where TAN is a common measure of general acidity. Increasing TAN, temperature, and flow velocity/turbulences and decreasing alloy contents of the bulk material accelerate naphthenic acid corrosion (NAC), which is often found to be very localized.⁹  This corrosion mechanism usually occurs at temperatures > 220°C and TAN > 0.5.¹⁵  The limit can be reduced to TAN > 0.1 for sweet, low sulfur crudes that have too little sulfur available to form a stable FeS scale.¹⁶  As a consequence, it was stated that TAN (and thus the overall acid content) is far too rough to evaluate the crude oil’s tendency to cause NAC.^17-18  Molybdenum contents > 2% in the bulk metal may restrict NAC corrosion rates,⁹  and it is argued that increasing sulfur contents can inhibit NAC as a result of FeS scale formation.¹⁹  Furthermore, the reactivity of each naphthenic acid, its distribution over boiling point range, and the influence of its molecular weight on its corrosion behavior are still under discussion.^18-19  For instance, experiments showed that naphthenic acids start to decompose at temperatures > 350°C.²  As a result, NAC is difficult to predict and thus observed frequently in CDU and HVU, particularly in furnace tubes and transfer lines between furnace and distillation column.¹⁶  Several pipe ruptures in CDU or HVU furnaces, for instance, were caused in the past by localized NAC. Those incidents can easily result in production losses of tens of million dollars.

The above chemical corrosion mechanisms can be further increased by erosion (by solid particles) or erosion-corrosion (e.g., for very high flow velocities and/or droplet impingement). Both mechanisms impair the FeS protective scale and accelerate sulfidation and naphthenic acid corrosion, in many cases very locally. Particularly, transfer lines are susceptible to these effects because of the presence of high flow velocities and two-phase flow. To control and limit high-temperature corrosion, materials and corrosion engineers usually consider one or several of the following approaches:

• •

  Materials selection (by applying, e.g., corrosion resistant alloys [CRAs])

• •

  Coatings (to protect the bulk material from the corrosive attack)

• •

  Corrosion allowance (by increasing nominal wall thickness to allow a certain degree of corrosion)

• •

  Corrosion inhibition (by injecting chemicals that, e.g., neutralize corrosive fluids)

• •

  Process adjustments (by limiting parameters such as sulfur content, TAN, or temperature)

• •

  Corrosion monitoring (by applying, e.g., non-destructive testing [NDT] to control the remaining wall thickness)

This paper focuses on the latter approach (corrosion monitoring).

There are different approaches in the petrochemical industry to monitor these corrosion mechanisms, where the most important ones are the following:

The use of corrosion coupons is the oldest and simplest method to estimate wall losses from corrosion.²⁰  A metallic corrosion coupon is weighed and introduced to the corrosive environment (e.g., inside of a pipe) where it remains for a certain time (e.g., some weeks) before being removed and re-weighed. From the difference in weight, an average corrosion rate can be calculated to obtain an estimate of the corrosion rate for the surrounding pipe/vessel.²¹  As a result of its simplicity, the method can be applied to many corrosion mechanisms and, thus, to different units of an oil refinery.²² 

Electrical resistance (ER) probes are also introduced into the process stream where they start to corrode. The resulting thickness reduction of the probe’s cross section is proportional to an increase in its electrical resistance, which is measured by an instrument outside of the pipe/vessel. This allows calculation of corrosion rates that are usually monitored online, which is a clear advantage in comparison to only a few values per year from weight loss coupons. Furthermore, ER probes can be applied to almost all corrosion environments²³  and they can have a high resolution down to tenths of nanometers.²⁴  They can be used, for instance, to evaluate the effectiveness of a chemical corrosion inhibitor in the atmospheric distillation overhead system.^3,25  The limits of the ER technique are, however, that they are intrusive and that the measurements are very local.²⁴  Also, ER probes have to be replaced when the maximum corrosion depth is exceeded (usually between 0.05 mm and 0.6 mm, dependent upon probe geometry²³ ).

Linear polarization resistance (LPR) is an electrochemical technique that allows determining corrosion rates in real time. Such intrusive probes usually consist of three electrodes that measure the electrochemical corrosion potential in corrosive environments. For this reason, the method can only be applied if an electrolyte (such as, e.g., water) is present, being restricted to electrochemical corrosion.²⁶  Hence, LPR cannot be used for corrosion monitoring in hydrocarbon streams in CDU and HVU where temperature is above the water boiling point.

Pulsed eddy current (PEC) probes were introduced in 2001²⁷  and are attached to the pipe’s external surface where they use a pulsed magnetic field. This generates eddy currents in the wall, which again induce voltages in the receiver coil of the PEC probe. This signal is used to calculate the remaining wall thickness, which is used as an indicator of corrosion, but it is not as accurate as, e.g., UT wall thickness measurements. PEC probes are portable and can thus be used at several corrosion monitoring locations (CMLs);²⁸  however, good accessibility for manual data collection is needed. The measuring system withstands metal temperatures up to 550°C but is restricted to conductive metals (carbon and low-alloyed steel)²⁹  and cannot be used as an online monitoring system.

Field signature measurements (FSM) involve a network of sensing pins or electrodes that are mounted on the external surface of a pipe or vessel producing an electrical field in the wall.²⁴  After an initial voltage measurement, subsequent changes in electrical field pattern are detected and compared against the initial measurement to detect local changes in wall thickness (accuracy: ±0.01 mm²⁴ ). The comparably large measuring area of FSM (e.g., an entire pipe elbow) allows detecting both general wall loss and localized corrosion. Owing to its complexity and thus elevated installation and operational costs, such measurements are usually restricted to a few locations within a chemical plant where corrosion hot spots are expected.²⁴ 

Non-intrusive UT measurements are the fastest method to reliably measure wall thickness and thus to monitor general wall loss. The wall thickness is calculated from the reflection of the ultrasonic signal at both external and internal surface. But the exact measuring location usually varies during periodic manual inspection, which limits accuracy of corrosion rates. To reduce this uncertainty, some systems monitor both measurement and location.³⁰  A further approach to increase the accuracy of several measurements at one location is to install permanent UT sensors that remain at the corresponding CML for a longer period such as months or even years.^5,8  In comparison to FSM, UT sensors provide local measurements as a result of their comparably low measuring area (usually some cm²), which makes detection of local corrosion phenomena difficult. An important advantage of UT sensors is that many more CMLs can be monitored by simple UT sensors for the same cost as by FSM.⁶ 

Table 1 summarizes pros and cons of the discussed corrosion monitoring methods, disclosing that UT sensors provide important advantages for corrosion monitoring in an oil refinery.

As a result of their suitability and flexibility, an increased number of ultrasonic sensors were installed to monitor high-temperature corrosion in CDU and HVU. All sensor locations were selected on the basis of three parameters:

• •

  Inspection history, particularly regarding former UT and radiographic testing (RT) measurements at the corresponding CMLs (to focus on corrosion hot spots);

• •

  Corrosion rates, calculated by a simulation tool that predicts high-temperature corrosion on the basis of pipe/vessel geometry, fluid, process conditions, and crude run history; and

• •

  Accessibility of each CML (to reduce installation and maintenance efforts).

As a result, pipe of the CDU and HVU systems discussed next were selected for the installation of such sensors.

Permasense^† UT sensors were mounted with their H-shaped foot on a pair of threaded studs that had been welded on the pipe’s external surface by a drawn-arc stud welding process (see Figure 1[a]). The challenge here was to do this while the pipe was in service at temperatures between 250°C and 410°C. Preceding welding procedure qualification report, welding procedure specification and selection and examination of specialized welders ensured maximum safety. This eventually allowed minimization of the turnaround scope. Furthermore, manual NDT measurements (UT + RT) were executed at each sensor location to check the actual wall thickness prior to welding. Ultrasonic sensors were installed, particularly at pipe elbow outlets (in the 80° position in Figure 1[b]). With increasing pipe diameter (from 150 mm to 800 mm), sensors were also attached at elbow inlet (10° position) and apex (45° position). As a result, many pipe elbows and some T-pieces of this CDU/HVU complex are now monitored continuously for corrosion. The pipe material was carbon steel at CMLs in loops 2, 3, and 4 and 5%Cr-0.5%Mo steel (ASTM A355, P5/X12CrMo5) at CMLs in loops 1, 5, and 6 (see Table 2).

To protect the electronics in the orange sensor head from the heat of the near pipe, heat shield plates were positioned on the sensor guide waves, which were attached afterward to the insulation. The measuring frequency of all sensors was chosen to be twice a day to allow increased battery life, also taking into account that such wall thickness measurements are usually mid- to long-term analyses.

The UT signal travels from the sensor head through two stainless steel guide waves to the sensor foot where it enters the pipe wall. It is then reflected particularly on both external (first echo) and internal (second echo) pipe surface. The corresponding UT waveform (or “A-scan”) is shown in Figure 2, emphasizing two important parameters that are recorded by the software for each sensor:

• •

  Wall thickness

  • ○

    Calculated by time-of-flight analysis from the interval between both main echoes that are detected by the “EP” (envelope peaks) approach;

  • ○

    Used to calculate corrosion rates (by trending wall thickness over time) as a measure of general wall loss at the corresponding CML.

• •

  Shape of back wall reflection

  • ○

    Compared to the back wall echo shape from previous and subsequent measurements (= change in shape of back wall reflection) and interpreted as a measure of change in roughness of the internal surface;

  • ○

    Used to calculate “PSI” (Permasense^† shape indicator) at the corresponding CML.

It is important to understand the difference between both parameters. Note that sulfidation usually leads first to a slight increase in surface roughness. This effects initiates at a very low order of magnitude such as micrometers or even nanometers. After some time, surface roughening accumulates and results in measurable wall thinning, from which one can calculate corrosion rates. The dimensionless parameter PSI was designed to describe the first stage (surface roughening) by focusing on the second echo of the UT waveform (see Figures 2 and 3). Accordingly, the sensor software analyzes the shape of this echo over time. Figure 3 shows an example of one sensor, revealing how significantly the echo’s shape can change (here: during 75 d). Remember that this second peak represents the appearance of the measuring area on the pipe’s internal surface (such as the first peak represents the external surface). Even small degrees of corrosion roughen the internal surface and change the shape of the second UT echo. Moreover, Figure 3 shows that wall thickness did not change in this example, which is shown by the distance between both main peaks that remained constant during 75 d. The shape of the back wall reflection is thus a parameter sensitive to detection of even low degrees of corrosion, and wall thickness can indicate elevated corrosion activity.

Each PSI value was calculated by the UT software by analyzing 60 different UT measurements/waveforms. With a measuring frequency of twice a day, all PSI values were thus based on data from 30 d. Furthermore, each PSI value was calculated from 30 preceding and 30 subsequent UT waveforms. This means that PSI data were always first available with a “delay” of 15 d (for a measuring frequency of twice a day). The exact algorithm used to compare the back wall reflection for different UT waveforms in order to calculate PSI was developed experimentally by the company in the last few years. It is unknown to the end-user and is thus not subject of this publication.

High PSI values are related to high changes in roughness indicating elevated corrosion on the internal pipe surface and vice-versa. Note that PSI is not influenced by process temperature because temperature does not influence the back wall signal’s shape. In contrast, wall thickness is influenced by temperature because the distance between both main peaks of the UT waveform increases with increasing temperature. This makes PSI an additional, very sensitive measuring value that complements conventional wall thickness readings.

It is of note that the sensor’s “footprint” (measuring area) is approximately 1 cm², which emphasizes that localized corrosion such as pitting cannot be detected reliably with such sensors. In other words, the sensor network should have a sufficient amount of sensors to allow reliable and repeatable measurements that show the potential for high-temperature corrosion in the whole pipe. Local effects owing to, e.g., turbulent flow or droplet impingement might only be evaluated with a very high number of sensors, which increases costs. On the other hand, only a few sensors are needed to monitor general wall loss in a comparably large area of a petrochemical plant.

After installation, the sensors automatically build a self-healing wifi network and connect to gateway (wifi antenna) and server. Software records UT waveform and wall thickness of each sensor, allowing further data processing to calculate corrosion rates and PSI. Thus, corrosion engineers can check these corrosion parameters at their office computers at any time, which makes the UT sensors a smart online corrosion monitoring tool. Furthermore, it should be noted that all sensors are wireless. Their intrinsically safe batteries can last for up to 7 y, dependent upon measuring frequency and stability of the sensor network. Also, these sensors withstand metal surface temperatures up to 600°C, which again highlights the flexibility of the measuring system.

Today, it is common practice that crude oil diets processed in oil refineries vary in very short time intervals as a result of the increasing need for crude flexibility during the last years (recall Introduction). In many cases, crude diets and thus parameters such as chemical composition, viscosity, flow conditions, or yields (in % per boiling point range) can change significantly within days or even hours. As a consequence, high-temperature corrosion phenomena in CDU and HVU vary over time and “corrosion hot spots” can change their severity and location. Hence, it is important to monitor crude parameters that contribute to high-temperature corrosion. In this study, the following crude parameters were obtained from the company’s crude assay database for the corrosion loops listed in Table 2:

• •

  Total sulfur content (determined according to ASTM D2622);

• •

  TAN (determined according to ASTM D974);

• •

  Mercaptan content (determined according to ASTM UOP163).

Figure 4(a) shows average values (9 month) of total sulfur content and TAN for six different CDU and HVU loops. Accordingly, sulfur content (0.82 wt%) and TAN (0.28 mg KOH/g) of the crude oil diet were comparably low. Both parameters can reach levels of several wt% (S content) or several mg KOH/g (TAN). This suggests that particularly sweet, low TAN crudes were processed during this period of 9 months. Furthermore, the diagram reveals a correlation between sulfur content and TAN. The mean sulfur/TAN ratio was measured to be > 2.0, except for kerosene (1.2) and heavy vacuum gas oil (HVGO, 1.7). Low ratios, < 1.0, can be interpreted as an indication of an increased susceptibility to NAC.¹⁶  Accordingly, increased NAC was found elsewhere in CDU and HVU of an oil refinery while processing very sweet, low TAN crude oils such as WAF (West African) crudes that can have sulfur/TAN ratios even below 0.5.¹⁶  As a result (from Figure 4[a]), sulfidation is expected to be the dominating corrosion mechanism (compared to NAC).

To give an overview of how such crude parameters vary over time, Figure 4(b) shows the mercaptan content during 10 months for the same CDU and HVU loops. In accordance with the total sulfur content (see Figure 4[a]), the mercaptan content was also found to be comparably low (< 25 ppm), indicating a low risk for mercaptan corrosion.¹³  It should be noted that there is a strong variation of mercaptans over time, but not for the total sulfur content. This emphasizes that each crude oil usually has a very different mercaptan concentration, independent of total sulfur. As known from experiments,¹³  Figure 4(b) shows elevated mercaptan contents for middle-distillates (such as gas oil) and low mercaptan contents for higher temperatures (here: LR [long (atmospheric) residue] and HVGO). This is a result of the boiling point range distribution of mercaptans and likely also a result of their decomposition at temperatures > 300°C.^13-14 

A typical plot of wall thickness vs. time (110 d) is shown by Figure 5(a) for one sensor that was installed in the kerosene system. The corresponding graph reveals very low wall thickness data scattering of ±0.01 mm. This is very accurate in comparison to manual UT thickness measurements where it is usually almost impossible to repeatedly measure at exactly the same position. Consequently, inaccuracy can increase by 1 to 2 magnitudes. The other curve in Figure 5(a) is temperature data that was measured with a nearby temperature sensor. It is obvious that both curves have a very similar shape, disclosing a strong dependence of measured wall thickness on process temperature. Accordingly, changes in temperature of only 3 K influence wall thickness readings significantly. This makes calculation of corrosion rates inaccurate, particularly if they are low. The corrosion rate indicated in Figure 5(a) (0.07 mm/y) corresponds to the slope of the linear fit of the wall thickness–time curve and was not compensated for temperature.

The propagation velocity of the UT signal depends upon metal temperature. Increasing temperatures reduce this velocity, increasing the distance between both main peaks in Figure 2. This is interpreted by the software as increasing wall thickness, which is, of course, not realistic. Furthermore, parameters such as insulation conditions, surrounding temperature, etc., are different at each sensor location, resulting in a different susceptibility of each sensor to temperature variations. As a consequence, “manual” or automatic compensation for temperature by the sensor software (by applying, e.g., a general correction factor) is difficult. Therefore, the sensor producer recently released a new sensor generation that includes a thermocouple. The software then compensates all wall thickness readings for temperature, which is intended for increasing the measuring system’s accuracy. The sensors of this study do not have such a thermocouple because they had been installed earlier.

Figure 5(b) shows the comparison between wall thickness and PSI for the same UT sensor. Interestingly, PSI data scatter very little. Furthermore, wall thickness and PSI developed very differently over time. Accordingly, PSI indicates a strong increase in corrosiveness between days 40 and 60, while the real wall thickness likely did not change significantly if one considers the temperature influence from Figure 5(a). A possible reason for this behavior (change in crude corrosiveness) is discussed in detail in the Correlation Between PSI and Crude Parameters section.

It should be noted here that changes in internal roughness (that are actually used to calculate PSI) can also slightly influence wall thickness readings. This was observed in a few cases where a small but sharp change in wall thickness by up to 0.1 mm was measured. This observation could not be explained by a change in temperature but by PSI, which was at maximum at that moment. Strong and sudden changes in internal roughness (high PSI) can smear the shape of the second UT echo, which modifies the exact peak location of this echo (recall Figure 3). This is interpreted by the software as slight but sharp change in wall thickness, which again affects overall corrosion rate. The extent of this effect was determined to be very low (< 5%) by comparing all wall thickness data to process temperature. Affected data was excluded and not considered for the calculation of corrosion rates.

The two main effects can be summarized as follows:

• •

  Process temperature (on wall thickness readings but not on PSI),

• •

  Internal roughness (on wall thickness readings and PSI).

Accordingly, PSI can be used as a first, very sensitive indicator of corrosion (in the form of changes in internal roughness), where wall thickness/corrosion rates indicate subsequent wall thinning restricted by temperature and, to some degree, by roughness. Also, data trending over time is more accurate for PSI than for corrosion rates.

In a second step, average values for corrosion rate and PSI data from 9 months were calculated and compared for six different corrosion loops (see Figure 6[a]). Note that all measured average corrosion rates (UT) were generally found to be low in comparison to rates that are typically observed in these CDU/HVU corrosion loops of an oil refinery. For gas oil reflux and long residue, the corrosion was even measured to be zero. The diagram compares measuring values to theoretical corrosion rates that were calculated from modified McConomy curves on the basis of Table 3. Therefore, temperature and material (recall Table 2) were used to determine corrosion rates for 0.5 wt% sulfur content, which was adjusted by a multiplier to obtain a theoretical corrosion rate for the corresponding sulfur content. Note that all corrosion rates in Figure 6(a) and Table 3 are normalized values, which means that absolute values (in mm/y) were transformed into relative values (in %).

As a result, no correlation was found between measurement (UT) and prediction (McConomy), see Figure 6(a). To explain this, it should be noted that hydrogen or H₂S were assumed to be absent (McConomy curves). Furthermore, this comparison points out that parameters such as type and distribution of sulfur compounds/naphthenic acids or flow conditions, etc., influence high-temperature sulfidation significantly. They are, however, not considered by the McConomy approach. This challenges the comparison of measured to theoretical corrosion rates. Moreover, it appears from Figure 6(a) that theoretical McConomy corrosion rates may be overly conservative for the corrosion observed in this study.

Figure 6(a) also shows PSI values that can range by definition from 255 (highest corrosion) to 0 (no corrosion). Hence, the presented PSI values are comparably low (< 100), which matches with the (low) level for UT corrosion rates in the same diagram. Furthermore, PSI correlates to the pipe material in Figure 6(a): PSI is lowest for loops 1, 5, and 6 in which 5 wt% Cr steel was used instead of carbon steel as in the other loops. As both PSI and UT corrosion rates are based on data from a large number of sensors, one can conclude that the crude oil diet processed in this period of time was not very corrosive in the high-temperature areas of CDU and HVU. Moreover, Figure 6(a) reveals significant differences between UT corrosion rates and PSI, although both are measuring data from UT sensors. It is probable that temperature (which influences wall thickness and thus corrosion rate, but not PSI) is the main reason behind this discrepancy. In other words, PSI values seem to show again a higher sensitivity to corrosion and thus an increased significance compared to corrosion rates.

An interesting comparison is given in Figure 6(b) that shows average UT corrosion rates and PSI for three different CML positions on pipe elbows (recall Figure 1[b]). It is a common belief that the elbow back side (45° position) is subject to the highest corrosive attack in each elbow because shear stresses resulting from flow impingement are expected to be the highest here. Figure 6(b) suggests a maximum in corrosion somewhere between 45° and 80° (elbow outlet). The lowest corrosion rates and PSI values were measured on the 10° position, which sounds reasonable. At this point, it should be highlighted that, e.g., computational fluid dynamics simulation can give a “corrosion profile” for such an elbow. This way, corrosion hot spots resulting from high local shear stresses that impair the passive FeS layer can be found.

As outlined earlier, important key factors for high-temperature corrosion in CDU/HVU are further process parameters such as flow velocity and pressure. An increase in these parameters can result in increased corrosion, owing to the increase of shear stresses at the internal pipe surface. This impairs the FeS protective layer, which facilitates corrosion of the bulk metal. Figure 7(a) discloses that no correlation was found between average values for flow velocity and corresponding PSI values. While PSI was highest in corrosion loops 2, 3, and 4, the corresponding flow velocity was very low (and vice-versa for the other loops). Similar results were found for the comparison of mean flow velocity with mean corrosion rates (not shown in Figure 7). This observation confirms that the UT sensors used here collect local measurements on the one hand, which likely do not fully represent the corrosion in the whole pipe. On the other hand, the flow regime usually affects corrosion locally. Examples are elbow back sides resulting from flow impingement or locations downstream of reinforced welds that cause local turbulences. In other words, flow velocity is a key factor but its effects on high-temperature corrosion cannot be detected reliably by local measurements.

Interestingly, pressure was observed to correlate to some degree to corrosion rate (see Figure 7[b]). Accordingly, corrosion rates were highest in loops with elevated pressure, such as the HVGO system. This suggests that pressure might be a key factor in sulfidation behavior, although, on the other hand, PSI did not show a clear influence by pressure (not shown in Figure 7).

Summarizing Figures 7(a) and (b), it should be pointed out that the UT sensors used here collect local measurements that have been averaged. Consequently, “local” influencing factors (such as flow velocity) seem to be dominated by “global” factors such as metallurgy, chemical composition (e.g., sulfur content), or pressure.

A very important question is if the UT measuring data correlates to crude parameters that are related to sulfidation and naphthenic acid corrosion. Therefore, measuring data was compared to crude parameters dependent upon time. As a result of the influence of temperature and roughness on corrosion rates (recall the Wall Thickness, Corrosion Rates, and PSI section), it was impossible to determine accurate corrosion rates over a small period of time, such as one month. The dependence on temperature variations and the resulting data scattering was simply too high. It was, however, possible to accurately average PSI values for each month. These mean values were compared to monthly mean values for the total sulfur content during the same 10 months period (see Figure 8). According to these six graphs (one per corrosion loop), there seems to exist a clear correlation between PSI and sulfur content. While both variables were found to have a minimum somewhere between months 4 and 5, there was a higher corrosion activity/sulfur content before and afterward (maximum in months 2 and 9).

Note that, for purposes of clarity, the vertical sulfur content axis of these six diagrams have different scales as a result of very different sulfur levels in each loop (recall Figure 4[a]). This makes the correlation between PSI and sulfur content quite significant, particularly for the kerosene loop (Figure 8[b]). Here, the total sulfur content was just some hundredths of wt%. The clear correlation of both curves thus emphasizes (1) the high sensitivity of the UT measuring system and (2) the relevance of PSI.

The crude oil’s total sulfur content covers many different sulfur species with very different tendencies for high-temperature corrosion. For this reason, it was often stated that such a “general” parameter would not be appropriate to predict sulfidation. This is challenged by the results in Figure 8. One should remember here that the total sulfur content is comparably easy to determine and to monitor in an oil refinery.

Two other influencing crude parameters (TAN and mercaptan content) are compared to PSI in different diagrams (see Figures 9 and 10). Note that the PSI curves in these two figures are consequently the same, as shown in Figure 8. Furthermore, the scales of all TAN axes in Figure 9 were adjusted to different typical TAN levels of each stream to improve clarity (as also done in Figure 8). As a result, no clear correlation was found between monthly TAN and PSI values for all systems (see Figure 9), but to some degree for the kerosene and gas oil loops (Figures 9[b] through [d]). As TAN is also a mixture of many acids with very different properties, it was suggested that TAN could not be used as an appropriate parameter to predict NAC.^9,16-18 

Some correlation was found between mercaptan content and PSI (see Figure 10) that shows mercaptan data from Figure 4(b). Interestingly, the highest correlation was observed for rather “cool” kerosene on the one hand, and high-temperature streams (crude oil, LR, and HVGO) on the other hand, where the mercaptan content was low (< 16 ppm). Mercaptans are commonly expected to decompose at temperatures between 300°C and 400°C,^13-14  which makes this observation even more interesting.

Table 4 summarizes the correlation between PSI and crude parameters (Figures 8 through 10) by comparing both curves of each diagram. The percentage of cases where both curves have a positive or negative slope between two neighbored months is given. From this overview, it becomes clear that both total sulfur content and TAN seem to be suitable to predict the corrosion that was measured with UT sensors used in this study, although the UT corrosion rates were found to be very low (recall Figure 6[a]). Accordingly, the total sulfur content correlates well with high-temperature corrosion in crude oil and middle distillates where TAN correlates well with corrosion in gas oil. The very low correlation of TAN with PSI values at high temperatures (LR and HVGO) is surprising, as TAN was measured here to be highest of all corrosion loops. Naphthenic acids are just some of many different acids that are considered in TAN calculation. It is difficult to predict which loop naphthenic acids concentrate in but they were already found to cause severe corrosion in HVUs.²  Furthermore, good correlation was found between mercaptan content and UT wall thickness measurements for crude oil, kerosene, LR, and HVGO pipe.

• •

  Two parameters were determined from continuous wall thickness measurements by wireless UT sensors that are installed permanently on piping in six different corrosion loops of an oil refinery’s CDU and HVU:

  • •

    Corrosion rates (as a measure of general wall thinning, derived from wall thickness);

  • •

    PSI (change in shape of UT back wall signal as a measure of change in internal roughness).

• •

  The results revealed two particular influencing factors:

  • •

    Temperature (that influences wall thickness and corrosion rates, but not PSI);

  • •

    Internal roughness (that can, when changing significantly, reduce wall thickness measuring accuracy).

• •

  Furthermore, the following results were observed:

  • •

    PSI seems to provide more accuracy in corrosion monitoring than wall thickness/corrosion rates, mainly because of uncertainty caused by temperature influence;

  • •

    No correlation between (measured) UT corrosion rates and (theoretical) McConomy corrosion rates;

  • •

    Clear correlation between pipe material and corrosion parameter PSI (low PSI for P5 material and elevated PSI for carbon steel piping);

  • •

    No correlation between measuring data and flow velocity but increased correlation between measuring data and process pressure.

• •

  Measuring data was compared to three crude parameters from crude run history: total sulfur content, TAN, and mercaptan content. The most important results from this analysis are as follows:

  • •

    Corrosion rates are too inaccurate to compare them to crude parameters for short periods of time such as, e.g., one month;

  • •

    Good correlation between PSI and crude parameters for comparably short periods of time (1 month), particularly regarding total sulfur content and TAN;

  • •

    Significant dependence of this correlation upon CDU/HVU corrosion loop.

• Thus, one important result from this study is the observation of correlations between measuring data (PSI) and crude parameters that represent the crude oil diet’s potential for high-temperature corrosion in CDU/HVU. This allows improved corrosion prediction based on measuring data, which is an important step toward safer and more economic processing of opportunity crude oils.

The authors would like to thank the sensor producer and partner companies for their outstanding support during commissioning and subsequent service of the UT sensors. Furthermore, they appreciate the support of all colleagues from Shell Rheinland Refinery and further Shell sites who contributed to this work.