CORROSION PREDICTION METHODS AND SYSTEMS

Abstract

A hybrid model for predicting corrosion in a system integrates a physics-based model developed using laboratory data and a machine-learning model developed using in-field data. Said hybrid model may be used, for example, in methods by: determining a physics-based measurement of corrosion using a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab-based measurements; determining a machine learning-based measurement of corrosion using a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field-based measurements; and applying an ensemble method to the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to yield an estimated measure of corrosion of the substrate. The hybrid model may be applied to corrosion mechanisms that occur in, for example, hydrocarbon transportation systems, hydrocarbon production systems, hydrocarbon refining systems, and alkylation systems.

Description

FIELD OF INVENTION

The present disclosure relates to predicting corrosion including, but not limited to, naphthenic acid and sulfur corrosion in hydrocarbon transportation and refinery systems, sulfuric acid corrosion in alkylation systems, and hydrogen sulfide and carbon dioxide corrosion in hydrocarbon production and transportation systems.

BACKGROUND

Corrosion can be a major problem in a variety of industrial systems and processes. Corrosion can cause deterioration of valves, gauges, storage containers, pipelines, and other equipment. Corrosion can also cause leaks with potentially large environmental and financial costs.

By way of nonlimiting example, in refining hydrocarbons, naphthenic acid and reactive sulfur compounds are two key species that contribute to high temperature corrosion (e.g., at about 400° F. to 800° F. or above). This process at high temperatures can be hypothesized by generic free radical reactions as shown by Equations 1-4 due to a lack of water at the liquid phase.

Fe+2RCOOH→Fe(RCOO)₂+H₂  Eq. 1

Fe+R₁SR₂→FeS+R₁R₂  Eq. 2

Fe+H₂S→FeS+H₂  Eq. 3

Fe(RCOO)₂+H₂S→FeS+2RCOOH  Eq. 4

Generally, naphthenic acid in crude hydrocarbons is quantified by a total acid number (TAN), which is the amount of potassium hydroxide in milligrams needed to neutralize one gram of the hydrocarbon.

Existing approaches used to predict naphthenic acid/sulfur corrosion and other types of corrosion take a fundamental approach that applies thermodynamics, kinetics, and transport phenomena. For example, a diffusion-adsorption-reaction mechanism is described in Peng Jin, et al. “Effect of sulfur compounds on formation of protective scales in naphthenic acid corrosion in non-turbulent flow,” in Corrosion Science, 131, 2018, pages 223-234. However, when compared to observed corrosion in refineries, these fundamental models can significantly over predict the amount or rate of corrosion. While this does mitigate component failure due to corrosion, over predictions also cause refineries to shut down complete or portions of the processes to replace components that still have a reasonable lifetime.

SUMMARY OF INVENTION

The present disclosure relates to predicting naphthenic acid/sulfur corrosion using a hybrid model that integrates a physics-based model developed using laboratory data and a machine-learning model developed using in-field data.

The present disclosure includes methods that comprise: determining, via the computing system, a physics-based measurement of corrosion using a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab-based measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) an operational parameter; determining, via the computing system, a machine learning-based measurement of corrosion using a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field-based measurements, wherein the machine-learning model correlates (a) the machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the operational parameters; and applying, via the computing system, an ensemble method to the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to yield a estimated measure of corrosion of the substrate.

The present disclosure includes methods that comprise: providing a hybrid model that correlates two or more operational parameters to an estimated measure of corrosion comprising: a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) two or more operational parameters; a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field measurements, wherein the machine-learning model correlates (a) a machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the two or more operational parameters; and an ensemble method that correlates (a) the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to (b) the estimated measure of corrosion; and simulating values or ranges of values in the hybrid model for a first operational parameter of the two or more operational parameters and the estimated measure of corrosion; and generating a value or range of values for a second operational parameter of the two or more operational parameters.

The present disclosure includes computing devices that comprise: a processor; a non-transitory, computer-readable medium coupled to the processor; and instructions provided to the non-transitory, computer-readable medium, wherein the instructions are executable by the processor to perform the forgoing methods.

The present disclosure includes computing devices that comprise: a processor; a non-transitory, computer-readable medium comprising a hybrid model that correlates one or more operational parameters to an estimated measure of corrosion; a non-transitory, computer-readable medium comprising instructions configured to accept inputs that include one or more operational parameters and/or an estimated measure of corrosion; and run the hybrid model to produce an output that includes one or more operational parameters and/or an estimated measure of corrosion that are not inputs wherein the hybrid model comprises: a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) the one or more operational parameters; a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field measurements, wherein the machine-learning model correlates (a) a machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the one or more operational parameters; and an ensemble method that correlates (a) the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to (b) the estimated measure of corrosion.

BRIEF DESCRIPTION OF THE DRAWINGS

The following figures are included to illustrate certain aspects of the disclosure, and should not be viewed as exclusive configurations. The subject matter disclosed is capable of considerable modifications, alterations, combinations, and equivalents in form and function, as will occur to those skilled in the art and having the benefit of this disclosure.

FIG. 1 illustrates a nonlimiting example method of applying a nonlimiting example hybrid model 100 of the present disclosure.

FIG. 2 illustrates a nonlimiting example method of applying a nonlimiting example hybrid model 200 of the present disclosure.

FIG. 3A is a plot of the residuals from the physics-based model in accordance with non-limiting Example 1 of the disclosure.

FIG. 3B is a plot of the residuals from the hybrid model in accordance with non-limiting Example 1 of the disclosure.

FIG. 4A is a plot of the residuals from the physics-based model in accordance with non-limiting Example 2 of the disclosure.

FIG. 4B is a plot of the residuals from the hybrid model in accordance with non-limiting Example 2 of the disclosure.

FIG. 5A is a plot of corrosion rate as a function of TAN for the two nonlimiting examples of FIGS. 1 and 2 for different substrate materials.

FIG. 5B is a plot of corrosion rate as a function of TRS for two nonlimiting examples of FIGS. 1 and 2 for the different substrate materials.

FIG. 5C is a plot of corrosion rate as a function of temperature (° F.) for two nonlimiting examples of FIGS. 1 and 2 for the different substrate materials.

FIG. 5D is a plot of corrosion rate as a function of fluid flow velocity for two nonlimiting examples of FIGS. 1 and 2 for the different substrate materials.

DETAILED DESCRIPTION

The present disclosure relates to predicting naphthenic acid/sulfur corrosion using a hybrid model that integrates a physics-based model developed using laboratory data and a machine-learning model developed using in-field data. While naphthenic acid/sulfur corrosion is discussed in more detail herein than other corrosion processes, the concepts of the present disclosure can be further extended to other corrosion processes, or material designs for catalysts, lubricants, and polymers, where sufficient in-field data is available for developing the machine-learning portion of the present disclosure.

As described above, current corrosion models can greatly over-predict the amount of corrosion occurring, which leads to over-engineered systems and too frequent system maintenance/component replacements. More accurate, less over-predictive models and methods for estimating corrosion have several advantages. For example, the level of corrosion resistance and corrosion resistance to specific compounds relates to the cost of the material used to make different components within a system (e.g., a refinery). Further, not all components within a system require the same level of corrosion resistance. Having a more accurate model may enable designers and operators to choose materials with the proper metallurgies for specific components at specific locations in a system to provide the necessary corrosion resistance without having to over-engineer the component and unnecessarily use more expensive materials.

In another example advantage, more accurate, less over-predictive methods may help complex systems like refineries plan risk-based inspections accordingly for preventative maintenance.

In yet another example advantage, the methods and systems described herein may expand the operating windows of various parameters in the systems. For example, in refineries a more accurate, less over-predictive method for estimating naphthenic acid/sulfur corrosion may allow for (a) using feedstocks that are more readily available but have a higher TAN/sulfur, (b) operating at higher temperatures and/or greater throughputs, and/or (c) reducing, if not eliminating, the use of corrosion inhibitors. Similar advantages may be seen in other systems that are concerned with corrosion by employing the methods described herein.

FIG. 1 illustrates a nonlimiting example method of applying a nonlimiting example hybrid model 100 of the present disclosure. The hybrid model 100 includes a physics-based model 104 and a machine learning-based model 108. The physics-based model 104 is developed based on a fundamental approach using thermodynamics, kinetics, and transport mechanisms related to corrosion. The physics-based model 104 primarily uses laboratory data to develop the algorithms within the physics-based model 104.

Generally, laboratory data explores a relationship between a measure of corrosion to one or more parameters or conditions of the experiment. As used herein, the term “laboratory data” is data collected under controlled conditions where a number of parameters are varied. The term “laboratory data” does not limit the location or size of the facility in which the data is collected. For example, in corrosion experiments, laboratory data relative to corrosion rate as a function of fluid composition or substrate composition is typically collected using coupons, which are small samples of the material under test. Other substrate configurations like pipes may be used for collecting other laboratory data like corrosion rate as a function of flow rate. Herein, the term “substrate” refers generally to the solid material where corrosion may occur no matter the configuration (e.g., coupon, foil, threads, fibers, pipes, joints, and the like) of that material.

Examples of parameters in laboratory data that may be correlated to a measure of corrosion include, but are not limited to, total acid number (TAN) of the fluid, composition of the TAN, total reactive sulfur (TRS) of the fluid, composition of the TRS, the origin of the fluid, the composition of the fluid, temperature of the fluid, fluid density, fluid velocity, a corrosion inhibitor concentration in the fluid, a corrosion inhibitor composition, composition of the substrate, configuration (or shape) of the substrate, the phases of the fluid, the phase behavior of the fluid, the absence or presence of scale on the substrate, the composition of said scale, the density of said scale, and the like, and any combination thereof. Again, TAN and TRS are specific to naphthenic corrosion. One skilled in the art will recognize other parameters that may be used when other fluids are considered. For example, when a corrosive fluid is water-based, the pH, the total dissolved solids, and/or the salinity may be useful parameters to consider.

As used herein, the term “measure of corrosion” refers generally to a metric by a level of corrosion can be described. Examples of measures of corrosion include, but are not limited to, an amount of corrosion, a percent of substrate corroded, a depth of corrosion into the substrate, a strength of the substrate, a corrosion rate, a relative rate of corrosion, and the like, and any combination thereof.

Referring back to FIG. 1, the physics-based model 104 correlates (a) a physics-based measure of corrosion 106 to (b) operational parameters 102 (illustrated here as four operational parameters 102a-d). The algorithms and other aspects of the physics-based model that form that correlation are developed using a fundamental approach and laboratory data.

The machine learning-based model 108, in contrast to the physics-based model 104, is developed primarily using in-field data. As used herein, the term “in-field data” is data collected during the operation of a system or portion thereof. The systems may be the same or similar to the system to which the hybrid model 100 is applied.

Examples of parameters in in-field data that may be correlated to a measure corrosion include, but are not limited to, TAN of the fluid, composition of the TAN, TRS of the fluid, composition of the TRS, the origin of the fluid, the composition of the fluid, temperature of the fluid, fluid density, fluid velocity, a corrosion inhibitor concentration in the fluid, a corrosion inhibitor composition, composition of the substrate, configuration (or shape) of the substrate, the phases of the fluid, the phase behavior of the fluid, the absence or presence of scale on the substrate, the composition of said scale, the density of said scale, and the like, and any combination thereof. Because the parameters under which operations are performed in systems may vary over the operation, the foregoing parameters may be more than a single value and represent the history of conditions a substrate (portion of the component) to which it was exposed. For example, the fluid flow rate through a component of a system may have been varied based on the production level in the system. The corrosion may be measured after the component is removed from the system. The in-field data that trains the machine learning-based model 108 may include the fluid flow rate history and extent of corrosion of the substrate over that history. In another example, the fluid composition may change based on feedstock availability. Accordingly, the in-field data would need the compositional history to which the substrate was exposed.

In-field data is collected with analytical components (e.g., sensors, spectrometers, chromatographs, and the like) placed throughout the system (e.g., a refinery). Said analytical components may be in-line with the process occurring in the system or separate from the process. Examples of analytical components include, but are not limited to, temperature sensors, pressure sensors, flow meters, gas chromatographs, liquid chromatographs, infra-red spectrometers, UV-visible spectrometers, nuclear magnetic resonance (NMR) spectrometers and/or imagers, 2-dimensional gas chromatographers, mass spectrometers, x-ray diffraction (XRD) instruments, and the like, and any combination thereof.

The machine learning-based model 108 is trained using in-field data to correlate (a) a machine learning-based measure of corrosion 110 to (b) operational parameters 102 and the physics-based measure of corrosion 106.

The machine learning-based model 108 may include algorithms that include polynomials, generalized linear model, elastic net, least absolute shrinkage and selection operator (Lasso), Ridge, boosting, extreme gradient boosting, support vector machine, neural networks, decision trees/random forest methods, kernal methods, Multivariate adaptive regression (MARS) methods, polyMARS methods, reinforcement learning methods, Gaussian process models, and the like, and any ensemble thereof. Examples of neural networks include, but are not limited to, perception, feed forward, radial basis, deep feed forward, recurrent neural network, long-short term memory, gated recurrent unit, auto encoder, variational auto encoder, denoising auto encoder, sparse auto encoder, Markov chain, Hopfield network, Boltzmann machine, restricted Boltzmann machine, deep belief network, deep convolutional network, deconvolutional network, deep convolutional inverse graphics network, generative adversarial network, liquid state machine, extreme learning machine, echo state network, deep residual network, Kohonen network, neural turning machine, and the like. Examples of kernal methods include, but are not limited to, kernel perceptron, Gaussian processes, principal components analysis, canonical correlation analysis, ridge regression, spectral clustering, linear adaptive filters, and the like.

Depending on the amount of in-field data, the machine learning-based model 108 may be trained with the shallow learning method (less in-field data) or deep learning methods (more in-field data).

Without being limited by theory, it is believed that the physics-based model 104 informs the hybrid model 100 of the way in which corrosion relates to parameters (e.g., the shape of the curve for amount of corrosion as a function of temperature and/or TAN) and the machine learning-based model 108 modulates the magnitude of said relationship in the hybrid model 100. To achieve this blending of the two models, an ensemble method 112 is applied to the physics-based measure of corrosion 106 and the machine learning-based measure of corrosion 110 to produce an estimated measure of corrosion 114. The physics-based model incorporates the correct physical trends with machine learning to improve the prediction accuracy. Hybrid modeling is particularly useful for complex systems where only partial fundamental understanding is available.

Examples of ensemble methods 112 include, but are not limited to, Bayes optimal classifying, bootstrap aggregating, boosting, Bayesian model averaging, bucket modeling, stacking, and the like, and any combination thereof.

While FIG. 1 illustrates the components 104, 108, 112 of the hybrid model 100 as separate blocks, one skilled in the art will recognize that the components 104, 108, 112 and algorithms thereof may be intermingled.

The hybrid model 100 can be applied in a variety of methods. As illustrated in FIG. 1, operational parameters 102 (illustrated here as four operational parameters 102a-d) are inputs to the hybrid model 100. The operational parameters 102 are used as inputs to the physics-based measure of corrosion 106 and the machine learning-based measure of corrosion 110 to produce the physics-based measure of corrosion 106 and the machine learning-based measure of corrosion 110, respectively. The ensemble method 112 is then applied to produce the estimated measure of corrosion 114 for the operational parameters 102 provided.

In this example method, the operational parameters 102 may include data from an operating system and/or simulated (or estimated) data. For example, the hybrid method 100 can be applied to a system or component thereof for monitoring the estimated measure of corrosion 114 at various points within the system. When the estimated measure of corrosion 114 reaches a threshold, an operator can be notified so that inspection, remediation, and/or replacement can occur. In another example, the hybrid method 100 can be applied to a system or component thereof for predicting the estimated measure of corrosion 114 for a future time if one or more of the operational parameters 102 were changed. In yet another example, the hybrid method 100 can be used to simulate a system or component thereof and the estimated measure of corrosion 114 under a variety of different operational parameters 102. This example can be used, for example, by designers to determine the corrosion at various points within a system to be built. Then, the operational parameters can be varied, for example, to give the various components the same approximate lifetime so that shutdowns can be performed less frequently.

FIG. 2 illustrates a nonlimiting example method of applying a nonlimiting example hybrid model 200 of the present disclosure. The reference numbers in FIG. 2 are 2## and refer to the same reference numbers 1## of FIG. 1 including descriptions of training and structure. FIG. 2 illustrates an alternate method to FIG. 1 as shown by the arrow directions.

In the example method of FIG. 2, the hybrid model 200 has inputs of a desired estimated measure of corrosion 214 and all but one of the operational parameters 102. More specifically, operational parameters 102a-c are inputs, and operational parameter 102d is an output. In this example, the hybrid model 200 provides a value or range of values for the operational parameter 102d based on the desired estimated measure of corrosion 214 and the other operational parameters 102a-c. While this example estimates only one of the operational parameters, the hybrid model could be applied to provide a value or range of values for more than one operation parameter when a desired estimated measure of corrosion and the other operational parameters are inputs.

In this example, the operational parameters 202a-c inputs may include data from an operating system and/or simulated (or estimated) data. For example, the hybrid method 200 can be applied to a system or component thereof for optimizing and/or adjusting an operational parameter 202d to adjust (e.g., increase or decrease) the corrosion (desired estimated measure of corrosion 214) of a component. Thus, if an operator wanted to decrease corrosion of a component, the hybrid model 200 could be applied to determine a reduced flow rate through said component that would provide the desired decrease in corrosion. In another example, the foregoing example may use measured data from the system for two of the operational parameters 202a-b inputs and a new target (so simulated) value for the third operation parameter 202c when determining what the value or range of values for the fourth operational parameter 202d should be to give the desired estimated measure of corrosion 214. In yet another example, the hybrid method 200 can be used to simulate a system or component thereof under a variety of different operational parameters 102a-c for a given desired estimated measure of corrosion 214 to determine the value or range of values for the fourth operational parameter 202d. This example can be used, for example, by designers to determine the material that a substrate should be made of to have a desired lifetime of the corresponding component while minimizing cost of said component.

FIGS. 1 and 2 are nonlimiting examples of methods for applying the hybrid models described herein. One skilled in the art will recognize that the hybrid models can be applied to the corrosion and operational parameters a variety of systems and components thereof.

The hybrid model application methods of FIGS. 1 and 2 and variations thereof can be applied in various scenarios.

By way of nonlimiting example, methods described herein may include applying a hybrid model described herein to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and repairing and/or replacing a component comprising the substrate based on the estimated measure of corrosion. In said methods, the substrate may be a portion of a system where the operational parameters inputs to the hybrid model comprise real-time operational parameter data of the system and, optionally, historical operational parameter data of the system.

In another nonlimiting example, methods described herein may include applying a hybrid model described herein to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and building a system or portion thereof comprising a component that comprises the substrate, wherein a composition of the substrate is chosen based on the estimated measure of corrosion. In such methods, the operational data inputs to the hybrid model may be simulated operational parameter data, real-time operational parameter data of the system and/or a comparable system, and/or historical operational parameter data of the system and/or a comparable system.

The hybrid models described herein may be applied to a variety of different corrosion mechanisms in a variety of different chemical processing and/or manufacturing facilities. Chemical processing and/or manufacturing facilities (also referred to herein as chemical processing/manufacturing facilities) are typically sites where a variety of chemical processes can occur such as producing, refining, synthesizing, formulating, blending, and/or storing chemicals (e.g., fuels such as gasoline, diesel, and kerosene; commodity and specialty chemicals such as olefins, aromatics, monomers, polymers, surfactants, dyes and pigments, and fertilizers; catalysts; and the like). Examples of such corrosion mechanisms and related systems include, but are not limited to, naphthenic acid and sulfur corrosion in hydrocarbon transportation and refinery systems; acid corrosion in refinery systems; sulfuric acid corrosion in alkylation systems; hydrogen sulfide and carbon dioxide corrosion in hydrocarbon production and transportation systems; and the like.

For example, localized corrosion also occurs on metal surfaces below deposits present on the metal surface, also referred to corrosion under deposits. Corrosion under deposits is prevalent through many systems and methods. Accordingly, the corrosive species (e.g., acid, sulfur, and the like) and the mechanism of corrosion may vary. However, given a system and the corrosive species present (which would be readily known by one skilled in the art), the physics-based model and the machine learning-based model and, consequently, the hybrid model described herein can be readily derived (or estimated or determined).

Corrosion under deposits can be one of the top causes for leaks in refining and chemical manufacturing systems. Current methods of estimating such corrosion are conservative by about two orders of magnitude, which again leads to over-engineering of systems and unnecessary down time for maintenance. In hydrocarbon refining, three places that are most susceptible to corrosion under deposits are the crude unit overhead, the fractionator overheat, and hydroprocessor effluent trains. While refining facilities differ, on average there are six of these units in some combination at refineries. By applying the hybrid model described herein to correlate the physics-based modeling and actual corrosion and failure data (the machine learning-based model) with operational parameters like wall thickness data, corrosion under deposits prediction may be improved, which results in improved reliability, reduced unplanned capacity loss, and better turn around planning.

In a nonlimiting example of the present disclosure, methods described herein may include refining a hydrocarbon feedstock (e.g., crude hydrocarbons, vacuum resid, and the like, and any combination thereof), wherein the feedstock or a downstream product and/or distillate thereof is the fluid; measuring the operational parameter in real-time; applying a hybrid model described herein using the real-time operational parameter data as at least one input to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and monitoring the estimated measure of corrosion over time. The hybrid model here may be based on one or more corrosion mechanisms or corrosive species present in refining.

In another nonlimiting example, methods described herein may include refining a feedstock (e.g., crude hydrocarbons, vacuum resid, and the like, and any combination thereof), wherein the feedstock or a downstream product and/or distillate thereof is the fluid; measuring the operational parameter in real-time; applying a hybrid model described herein using the real-time operational parameter data as at least one input to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and changing a composition of the feedstock based on the estimated measure of corrosion over time. The hybrid model here may be based on one or more corrosion mechanisms or corrosive species present in refining.

In another nonlimiting example, methods described herein may include refining a feedstock (e.g., crude hydrocarbons, vacuum resid, and the like, and any combination thereof), wherein the feedstock or a downstream product and/or distillate thereof is the fluid; measuring the operational parameter in real-time; applying a hybrid model described herein using the real-time operational parameter data as at least one input to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and projecting the estimated measure of corrosion based on a change to the operational parameter. The hybrid model here may be based on one or more corrosion mechanisms or corrosive species present in refining.

In yet another nonlimiting example, methods described herein may include applying a hybrid model described herein to yield a value or range of values for a composition of a fluid that is a feedstock or a downstream product and/or distillate thereof (e.g., the method of FIG. 2 or a variation thereof where the composition of the fluid is operational parameter 102d); and sourcing a feedstock (e.g., crude hydrocarbons, vacuum resid, and the like, and any combination thereof) for the system based on the value or range of values for a second operational parameter. The hybrid model here may be based on one or more corrosion mechanisms or corrosive species present in refining.

Another corrosion/system example to which the hybrid model methods and systems may be applied is sulfuric acid corrosion in alkylation systems. Without being limited by theory, it is believed that the corrosion rate of sulfuric acid in an alkylation system is limited by the removal iron from a metal substrate as Fe₂SO₄, which is the reaction product of sulfuric acid and iron. Alkylation units require rigorous inspection due to the presence of sulfuric acid in the feedstocks for the alkylation process. The hybrid model can be applied to this corrosion/system example with similar advantages as refining systems to reduce over-engineering of the systems and mitigate unnecessary down time by adjusting operational parameters like temperature, flow rate, pressure, and the like as well as the source and/or composition of the feedstock.

In another nonlimiting example of the present disclosure, methods described herein may include reacting an olefin feed (e.g., butene, isobutene, propene, propane, n-butane, isobutane, and the like, and any combination thereof) in the presence of a catalyst (e.g., via an alkylation mechanism), wherein the olefin feed is the fluid; measuring the operational parameter in real-time; applying a hybrid model described herein using the real-time operational parameter data as at least one input to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and monitoring the estimated measure of corrosion over time. The hybrid model here may be based on one or more corrosion mechanisms or corrosive species present in alkylation systems.

In another nonlimiting example, methods described herein may include reacting an olefin feed (e.g., butene, isobutene, propene, propane, n-butane, isobutane, and the like, and any combination thereof) in the presence of a catalyst (e.g., via an alkylation mechanism), wherein the olefin feed is the fluid; measuring the operational parameter in real-time; applying a hybrid model described herein using the real-time operational parameter data as at least one input to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and changing a composition of the feedstock based on the estimated measure of corrosion over time. The hybrid model here may be based on one or more corrosion mechanisms or corrosive species present in alkylation systems.

In another nonlimiting example, methods described herein may include reacting an olefin feed (e.g., butene, isobutene, propene, propane, n-butane, isobutane, and the like, and any combination thereof) in the presence of a catalyst (e.g., via an alkylation mechanism), wherein the olefin feed is the fluid; measuring the operational parameter in real-time; applying a hybrid model described herein using the real-time operational parameter data as at least one input to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and projecting the estimated measure of corrosion based on a change to the operational parameter. The hybrid model here may be based on one or more corrosion mechanisms or corrosive species present in alkylation systems.

In yet another nonlimiting example, methods described herein may include applying a hybrid model described herein to yield a value or range of values for a composition of a fluid that is an olefin feed (e.g., the method of FIG. 2 or a variation thereof where the composition of the fluid is operational parameter 102d); and sourcing an olefin feed for the system based on the value or range of values for a second operational parameter. The hybrid model here may be based on one or more corrosion mechanisms or corrosive species present in alkylation systems.

Yet another corrosion/system example to which the hybrid model methods and systems may be applied is hydrogen sulfide and/or carbon dioxide (or carbonic acid) corrosion in hydrocarbon production and/or transportation systems.

In another nonlimiting example of the present disclosure, methods described herein may include producing the fluid and/or transporting the fluid (e.g., a mixture of hydrocarbons and corrosive species); measuring the operational parameter in real-time; applying a hybrid model described herein using the real-time operational parameter data as at least one input to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and monitoring the estimated measure of corrosion over time.

In another nonlimiting example, methods described herein may include producing the fluid and/or transporting the fluid (e.g., a mixture of hydrocarbons and corrosive species); measuring the operational parameter in real-time; applying a hybrid model described herein using the real-time operational parameter data as at least one input to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and changing a composition of the feedstock based on the estimated measure of corrosion over time.

In another nonlimiting example, methods described herein may include producing the fluid and/or transporting the fluid (e.g., a mixture of hydrocarbons and corrosive species); measuring the operational parameter in real-time; applying a hybrid model described herein using the real-time operational parameter data as at least one input to yield an estimated measure of corrosion of the substrate (e.g., the method of FIG. 1 or a variation thereof); and projecting the estimated measure of corrosion based on a change to the operational parameter.

The hybrid modeling of the present disclosure can also be applied to expedite the development of new catalysts, lubricants, and polymers. The present physics-based models use rules derived from fundamental understanding of the composition, structure, activity, performance, and the like of the corresponding catalysts, lubricants, and polymers. However, abundant data is available that can be applied to enhance the fundamental understanding via hybridization techniques. The resulting improved fundamental understanding may reduce the development cycles for new catalysts, lubricants, and polymers.

The methods described herein can, and in many embodiments must, be performed using computing devices or processor-based devices. “Computer-readable medium” or “non-transitory, computer-readable medium,” as used herein, refers to any non-transitory storage and/or transmission medium that participates in providing instructions to a processor for execution. Such a medium may include, but is not limited to, non-volatile media and volatile media. Non-volatile media includes, for example, NVRAM, or magnetic or optical disks. Volatile media includes dynamic memory, such as main memory. Common forms of computer-readable media include, for example, a floppy disk, a flexible disk, a hard disk, an array of hard disks, a magnetic tape, or any other magnetic medium, magneto-optical medium, a CD-ROM, a holographic medium, any other optical medium, a RAM, a PROM, and EPROM, a FLASH-EPROM, a solid-state medium like a memory card, any other memory chip or cartridge, or any other tangible medium from which a computer can read data or instructions. When the computer-readable media is configured as a database, it is to be understood that the database may be any type of database, such as relational, hierarchical, object-oriented, and/or the like. Accordingly, the embodiments of the present systems and methods may be considered to include a tangible storage medium or tangible distribution medium and prior art-recognized equivalents and successor media, in which the software implementations embodying the present techniques are stored.

The methods described herein may be performed using computing devices or processor-based devices that include a processor; a non-transitory, computer-readable medium coupled to the processor; and instructions provided to the non-transitory, computer-readable medium, wherein the instructions are executable by the processor to perform the methods described herein (such as computing or processor-based devices may be referred to generally by the shorthand “computer”). For example, a system may comprise: a processor; a non-transitory, computer-readable medium coupled to the processor; and instructions provided to the non-transitory, computer-readable medium, wherein the instructions are executable by the processor to perform a method comprising: determining, via the computing system, a physics-based measurement of corrosion using a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab-based measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) an operational parameter; determining, via the computing system, a machine learning-based measurement of corrosion using a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field-based measurements, wherein the machine-learning model correlates (a) the machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the operational parameters; and applying, via the computing system, an ensemble method to the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to yield an estimated measure of corrosion of the substrate. In another example, for example, a system may comprise: a processor; a non-transitory, computer-readable medium coupled to the processor and comprising a hybrid model that correlates two or more operational parameters to an estimated measure of corrosion comprising: a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) two or more operational parameters; a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field measurements, wherein the machine-learning model correlates (a) a machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the two or more operational parameters; and an ensemble method that correlates (a) the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to (b) the estimated measure of corrosion; and instructions provided to the non-transitory, computer-readable medium, wherein the instructions are executable by the processor to perform a method for predicting corrosion comprising: simulating values or ranges of values in the hybrid model for a first operational parameter of the two or more operational parameters and the estimated measure of corrosion; and generating a value or range of values for a second operational parameter of the two or more operational parameters.

Example Embodiments

A first nonlimiting embodiment of the present disclosure is a method comprising: determining, via the computing system, a physics-based measurement of corrosion using a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab-based measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) an operational parameter; determining, via the computing system, a machine learning-based measurement of corrosion using a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field-based measurements, wherein the machine-learning model correlates (a) the machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the operational parameters; and applying, via the computing system, an ensemble method to the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to yield an estimated measure of corrosion of the substrate. The first nonlimiting embodiment may further include one or more of: Element 1: wherein the operational parameter is selected from the group consisting of: a total acid number (TAN) of the fluid, a composition of the TAN, a total reactive sulfur (TRS) of the fluid, a composition of the TRS, an origin of the fluid, a composition of the fluid, a temperature of the fluid, a fluid density, a fluid velocity, a corrosion inhibitor concentration in the fluid, a corrosion inhibitor composition, composition of the substrate, configuration of the substrate, phases of the fluid, a phase behavior of the fluid, an absence or presence of scale on the substrate, a composition of said scale, a density of said scale, and any combination thereof; Element 2: the method further comprising: repairing and/or replacing a component comprising the substrate based on the estimated measure of corrosion; Element 3: the method further comprising: building a system or portion thereof comprising a component that comprises the substrate, wherein a composition of the substrate is chosen based on the estimated measure of corrosion; Element 4: the method further comprising: measuring the operational parameter in real-time; and monitoring the estimated measure of corrosion over time; Element 5: the method further comprising: measuring the operational parameter in real-time; and changing a composition of the fluid or an upstream fluid based on the estimated measure of corrosion over time; Element 6: the method further comprising: projecting the estimated measure of corrosion based on a change to the operational parameter; Element 7: wherein the operation parameter is a hydrocarbon transportation operational parameter, and wherein the substrate is a component or portion thereof of a hydrocarbon transportation system; Element 8: the method further comprising: transporting a hydrocarbon (e.g., through a pipeline), wherein the hydrocarbon is the fluid; Element 9: wherein the operation parameter is a hydrocarbon refining operational parameter, and wherein the substrate is a component or portion thereof of a hydrocarbon refining system; Element 10: the method further comprising: refining a feedstock, wherein the feedstock or a downstream product and/or distillate thereof is the fluid; Element 11: wherein the operation parameter is a hydrocarbon production operational parameter, and wherein the substrate is a component or portion thereof of a hydrocarbon production system; Element 12: the method further comprising: producing a hydrocarbon (e.g., from a subterranean formation), wherein the hydrocarbon is the fluid; Element 13: wherein the operation parameter is an alkylation operational parameter, and wherein the substrate is a component or portion thereof of an alkylation system; and Element 14: the method further comprising: reacting an olefin feed in the presence of a catalyst, wherein the olefin feed is the fluid. Examples of combinations include, but are not limited to, Element 1 in combination with one or more of Elements 2-6; two or more of Elements 2-6 in combination; Element 7 and/or Element 8 in combination with one or more of Elements 2-6, optionally in further combination with Element 1; Element 9 and/or Element 10 in combination with one or more of Elements 2-6, optionally in further combination with Element 1; Element 11 and/or Element 12 in combination with one or more of Elements 2-6, optionally in further combination with Element 1; Element 13 and/or Element 14 in combination with one or more of Elements 2-6, optionally in further combination with Element 1.

A second nonlimiting example embodiment is a computing system comprising: a processor; a non-transitory, computer-readable medium coupled to the processor; and instructions provided to the non-transitory, computer-readable medium, wherein the instructions are executable by the processor to perform the method of the first nonlimiting embodiment optionally including one or more of Elements 1-14 (e.g., in any of the foregoing combinations).

A third nonlimiting example embodiment is a method for predicting corrosion comprising: providing a hybrid model that correlates two or more operational parameters to an estimated measure of corrosion comprising: a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) two or more operational parameters; a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field measurements, wherein the machine-learning model correlates (a) a machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the two or more operational parameters; and an ensemble method that correlates (a) the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to (b) the estimated measure of corrosion; and simulating values or ranges of values in the hybrid model for a first operational parameter of the two or more operational parameters and the estimated measure of corrosion; and generating a value or range of values for a second operational parameter of the two or more operational parameters. The third nonlimiting example embodiment may further include one or more of: Element 1; Element 15: wherein the second operational parameter is a composition of the substrate, wherein the substrate corresponds to a component in a system, and wherein the method further comprises: using the component having the composition in the system; Element 16: wherein the second operational parameter is a composition of the fluid, wherein the substrate corresponds to a component in a system, and wherein the method further comprises: sourcing a feedstock for the system based on the value or range of values for a second operational parameter; Element 17: wherein the operation parameters are hydrocarbon transportation operational parameters, and wherein the substrate is a component or portion thereof of a hydrocarbon transportation system; Element 18: the method further comprising: transporting a hydrocarbon (e.g., through a pipeline), wherein the hydrocarbon is the fluid; Element 19: the operation parameters are hydrocarbon refining operational parameters, and wherein the substrate is a component or portion thereof of a hydrocarbon refining system; Element 20: the method further comprising: refining a feedstock, wherein the feedstock or a downstream product and/or distillate thereof is the fluid; Element 21: the operation parameters are hydrocarbon production operational parameters, and wherein the substrate is a component or portion thereof of a hydrocarbon production system; Element 22: the method further comprising: producing a hydrocarbon (e.g., from a subterranean formation), wherein the hydrocarbon is the fluid; Element 23: the operation parameters are alkylation operational parameters, and wherein the substrate is a component or portion thereof of an alkylation system; and Element 24: the method further comprising: reacting an olefin feed in the presence of a catalyst, wherein the olefin feed is the fluid. Examples of combinations include, but are not limited to, Element 1 in combination with Element 15 and/or Element 16; Element 17 and/or Element 18 in combination with Element 15 and/or Element 16, optionally in further combination with Element 1; Element 19 and/or Element 20 in combination with Element 15 and/or Element 16, optionally in further combination with Element 1; Element 21 and/or Element 22 in combination with Element 15 and/or Element 16, optionally in further combination with Element 1; Element 23 and/or Element 24 in combination with Element 15 and/or Element 16, optionally in further combination with Element 1.

A fourth nonlimiting example embodiment is a computing system comprising: a processor; a non-transitory, computer-readable medium coupled to the processor; and instructions provided to the non-transitory, computer-readable medium, wherein the instructions are executable by the processor to perform the method of the first nonlimiting embodiment optionally including one or more of Elements 1 and 15-24 (e.g., in any of the foregoing combinations), wherein the provided hybrid model is on non-transitory, computer-readable medium.

A fifth nonlimiting example embodiment is a computing system comprising: a processor; a non-transitory, computer-readable medium comprising a hybrid model that correlates one or more operational parameters to an estimated measure of corrosion; a non-transitory, computer-readable medium comprising instructions configured to accept inputs that include one or more operational parameters and/or an estimated measure of corrosion; and run the hybrid model to produce an output that includes one or more operational parameters and/or an estimated measure of corrosion that are not inputs wherein the hybrid model comprises: a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) the one or more operational parameters; a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field measurements, wherein the machine-learning model correlates (a) a machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the one or more operational parameters; and an ensemble method that correlates (a) the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to (b) the estimated measure of corrosion. The fifth nonlimiting example embodiment may include one or more of: Element 1; Element 25: wherein the output comprises a value or range of values for an operational parameter that is not an input; Element 26: wherein the output comprises the estimated measure of corrosion that is not an input. Said computing system (optionally including one or more of Elements 1, 25, and 26) may be a portion of a hydrocarbon transportation system, a hydrocarbon refinery system, hydrocarbon production system, or an alkylation system.

Unless otherwise indicated, all numbers expressing quantities of ingredients, properties such as molecular weight, reaction conditions, and so forth used in the present specification and associated claims are to be understood as being modified in all instances by the term “about.” Accordingly, unless indicated to the contrary, the numerical parameters set forth in the following specification and attached claims are approximations that may vary depending upon the desired properties sought to be obtained by the incarnations of the present inventions. At the very least, and not as an attempt to limit the application of the doctrine of equivalents to the scope of the claim, each numerical parameter should at least be construed in light of the number of reported significant digits and by applying ordinary rounding techniques.

One or more illustrative incarnations incorporating one or more invention elements are presented herein. Not all features of a physical implementation are described or shown in this application for the sake of clarity. It is understood that in the development of a physical embodiment incorporating one or more elements of the present invention, numerous implementation-specific decisions must be made to achieve the developer's goals, such as compliance with system-related, business-related, government-related and other constraints, which vary by implementation and from time to time. While a developer's efforts might be time-consuming, such efforts would be, nevertheless, a routine undertaking for those of ordinary skill in the art and having benefit of this disclosure.

While compositions and methods are described herein in terms of “comprising” various components or steps, the compositions and methods can also “consist essentially of” or “consist of” the various components and steps.

To facilitate a better understanding of the embodiments of the present invention, the following examples of preferred or representative embodiments are given. In no way should the following examples be read to limit, or to define, the scope of the invention.

EXAMPLES

A physics-based model for naphthenic acid/sulfur corrosion was produced by known methods using laboratory data collected over several years. The physics-based model was based on a diffusion-adsorption-free radical reaction pathway for naphthenic acid/sulfur corrosion.

A hybrid model for naphthenic acid/sulfur corrosion was produced using the foregoing physics-based model and a machine learning-based model trained with in-field data. The in-field data was from seven different ExxonMobil refineries. The machine learning-based model used extreme gradient boosting. Further, the ensemble method of the machine learning-based model was stacking.

Example 1. 52 historical data points from a first ExxonMobil refinery was used to validate and compare the physics-based model and the hybrid model. The operational parameters were input to the models, and the estimated measure of corrosion was output. In this instance, the measure of corrosion used was corrosion rate reported as mils per year (mil/yr). The estimated measure of corrosion was compared to the observed measure of corrosion by subtracting the measured from the estimated to give the residuals of each model. Negative residuals indicate where a model would underestimate the measure of corrosion. The room mean squared error (RMSE), or standard error, in the residuals that is calculated according to Eq. 1, where N is the number of data points.

$\mathrm{RMSE}=\sqrt{\frac{{\sum }_1^N{\left(\mathrm{predicted}-\mathrm{observed}\right)}^2}{N}}$

FIG. 3A is a plot of the residuals from the physics-based model, and FIG. 3B is a plot of the residuals from the hybrid model. Each model had only one data point where underestimation of corrosion would have occurred. As illustrated in the graphs, the physics-based model had more overestimation of corrosion than the hybrid model, which would translate to costly shut downs more often for maintenance/component replacement. This is also illustrated with the standard error of the physics-based model being 32 mil/yr, while the hybrid model was only 19 mil/yr, a 40% reduction in overestimation.

Example 2. 21 historical data points from a second ExxonMobil refinery was used to validate and compare the physics-based model and the hybrid model. Validation was performed as described in Example 1.

FIG. 4A is a plot of the residuals from the physics-based model, and FIG. 4B is a plot of the residuals from the hybrid model. The physics-based model and the hybrid model had one data point where underestimation of corrosion would have occurred. As illustrated in the graphs, the physics-based model had much more overestimation of corrosion than the hybrid model, which would translate to costly shut downs more often for maintenance/component replacement. This is also illustrated with the standard error of the physics-based model being 172 mil/yr, while the hybrid model was only 77 mil/yr, a 55% reduction in overestimation.

Examples 1 and 2 illustrate that the hybrid methods described herein reduce the overestimation of corrosion in systems while introducing little to no additional underestimation in corrosion.

Example 3. Using the two models, the corrosion rate of various substrates can be estimated as a function of various operational parameters. In this example, 5Cr0.5Mo steel (5CR), 9CrlMo steel (9CR), and commercial steel (CS) were used for the substrate materials.

FIG. 5A is a plot of corrosion rate as a function of TAN for the two models for the different substrate materials. FIG. 5B is a plot of corrosion rate as a function of TRS for the two models for the different substrate materials. FIG. 5C is a plot of corrosion rate as a function of temperature (° F.) for the two models for the different substrate materials. FIG. 5D is a plot of corrosion rate as a function of fluid flow velocity for the two models for the different substrate materials.

Therefore, the present invention is well adapted to attain the ends and advantages mentioned as well as those that are inherent therein. The particular examples and configurations disclosed above are illustrative only, as the present invention may be modified and practiced in different but equivalent manners apparent to those skilled in the art having the benefit of the teachings herein. Furthermore, no limitations are intended to the details of construction or design herein shown, other than as described in the claims below. It is therefore evident that the particular illustrative examples disclosed above may be altered, combined, or modified and all such variations are considered within the scope and spirit of the present invention. The invention illustratively disclosed herein suitably may be practiced in the absence of any element that is not specifically disclosed herein and/or any optional element disclosed herein. While compositions and methods are described in terms of “comprising,” “containing,” or “including” various components or steps, the compositions and methods can also “consist essentially of” or “consist of” the various components and steps. All numbers and ranges disclosed above may vary by some amount. Whenever a numerical range with a lower limit and an upper limit is disclosed, any number and any included range falling within the range is specifically disclosed. In particular, every range of values (in the form, “from about a to about b,” or, equivalently, “from approximately a to b,” or, equivalently, “from approximately a-b”) disclosed herein is to be understood to set forth every number and range encompassed within the broader range of values. Also, the terms in the claims have their plain, ordinary meaning unless otherwise explicitly and clearly defined by the patentee. Moreover, the indefinite articles “a” or “an,” as used in the claims, are defined herein to mean one or more than one of the element that it introduces.

Claims

1. A method comprising:

determining, via the computing system, a physics-based measurement of corrosion using a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab-based measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) an operational parameter;

determining, via the computing system, a machine learning-based measurement of corrosion using a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field-based measurements, wherein the machine-learning model correlates (a) the machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the operational parameters; and

applying, via the computing system, an ensemble method to the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to yield an estimated measure of corrosion of the substrate.

2. The method of claim 1, wherein the operational parameter is selected from the group consisting of: a total acid number (TAN) of the fluid, a composition of the TAN, a total reactive sulfur (TRS) of the fluid, a composition of the TRS, an origin of the fluid, a composition of the fluid, a temperature of the fluid, a fluid density, a fluid velocity, a corrosion inhibitor concentration in the fluid, a corrosion inhibitor composition, composition of the substrate, configuration of the substrate, phases of the fluid, a phase behavior of the fluid, an absence or presence of scale on the substrate, a composition of said scale, a density of said scale, and any combination thereof.

3. The method of claim 1 further comprising:

repairing and/or replacing a component comprising the substrate based on the estimated measure of corrosion.

4. The method of claim 1 further comprising:

building a system or portion thereof comprising a component that comprises the substrate, wherein a composition of the substrate is chosen based on the estimated measure of corrosion.

5. The method of claim 1 further comprising:

refining a feedstock, wherein the feedstock or a downstream product and/or distillate thereof is the fluid;

measuring the operational parameter in real-time; and

monitoring the estimated measure of corrosion over time.

6. The method of claim 1 further comprising:

refining a feedstock, wherein the feedstock or a downstream product and/or distillate thereof is the fluid;

measuring the operational parameter in real-time; and

changing a composition of the feedstock based on the estimated measure of corrosion over time.

7. The method of claim 1 further comprising:

refining a feedstock, wherein the feedstock or a downstream product and/or distillate thereof is the fluid; and

projecting the estimated measure of corrosion based on a change to the operational parameter.

8. A method for predicting corrosion comprising:

providing a hybrid model that correlates two or more operational parameters to an estimated measure of corrosion comprising: a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) two or more operational parameters; a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field measurements, wherein the machine-learning model correlates (a) a machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the two or more operational parameters; an ensemble method that correlates (a) the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to (b) the estimated measure of corrosion;

simulating values or ranges of values in the hybrid model for a first operational parameter of the two or more operational parameters and the estimated measure of corrosion; and

generating a value or range of values for a second operational parameter of the two or more operational parameters.

9. The method of claim 8, wherein at least one of the two or more operational parameters are selected from the group consisting of: a total acid number (TAN) of the fluid, a composition of the TAN, a total reactive sulfur (TRS) of the fluid, a composition of the TRS, an origin of the fluid, a composition of the fluid, a temperature of the fluid, a fluid density, a fluid velocity, a corrosion inhibitor concentration in the fluid, a corrosion inhibitor composition, composition of the substrate, configuration of the substrate, phases of the fluid, a phase behavior of the fluid, an absence or presence of scale on the substrate, a composition of said scale, a density of said scale, and any combination thereof.

10. The method of claim 8, wherein the second operational parameter is a composition of the substrate, wherein the substrate corresponds to a component in a system, and wherein the method further comprises:

using the component having the composition in the system.

11. The method of claim 8, wherein the second operational parameter is a composition of the fluid, wherein the substrate corresponds to a component in a system, and wherein the method further comprises:

sourcing a feedstock for the system based on the value or range of values for a second operational parameter.

12. A computing system comprising: a processor; a non-transitory, computer-readable medium comprising a hybrid model that correlates one or more operational parameters to an estimated measure of corrosion; a non-transitory, computer-readable medium comprising instructions configured to accept inputs that include one or more operational parameters and/or an estimated measure of corrosion; and run the hybrid model to produce an output that includes one or more operational parameters and/or an estimated measure of corrosion that are not inputs wherein the hybrid model comprises:

a physics-based model for a fluid's corrosion of a substrate based, at least in part on, lab measurements, wherein the physics-based model correlates (a) the physics-based measurement of corrosion to (b) the one or more operational parameters;

a machine learning-based model for the fluid's corrosion of the substrate based, at least in part on, in-field measurements, wherein the machine-learning model correlates (a) a machine learning-based measurement of corrosion to (b) the physics-based measurement of corrosion and the one or more operational parameters; and

an ensemble method that correlates (a) the physics-based measurement of corrosion and the machine learning-based measurement of corrosion to (b) the estimated measure of corrosion.

13. The computing system of claim 12, wherein the output comprises a value or range of values for an operational parameter that is not an input.

14. The computing system of claim 12, wherein the output comprises the estimated measure of corrosion that is not an input.

15. The computing system of claim 12, wherein at least one of the one or more operational parameters are selected from the group consisting of: a total acid number (TAN) of the fluid, a composition of the TAN, a total reactive sulfur (TRS) of the fluid, a composition of the TRS, an origin of the fluid, a composition of the fluid, a temperature of the fluid, a fluid density, a fluid velocity, a corrosion inhibitor concentration in the fluid, a corrosion inhibitor composition, composition of the substrate, configuration of the substrate, phases of the fluid, a phase behavior of the fluid, an absence or presence of scale on the substrate, a composition of said scale, a density of said scale, and any combination thereof.