A review of research methods for corrosion under insulation

1 Introduction

Insulations are commonly applied to cover specific equipment for temperature maintenance, personnel protection, fire protection, sound control, condensation control, and process control in all the refining, industrial, power, petrochemical, onshore, and offshore industries (NACE International 2010). Corrosion under insulation (CUI) has become a major common problem suffered by all these industries worldwide. It refers to the external corrosion of metallic components and equipment under external cladding or jacketed insulation (Landtsheer 2020). While some incidences of CUI were first reported in the 1950s, people did not pay attention and started to mitigate CUI until the 1980s. The World Corrosion Organization estimated that the corrosion costs were around 2.2 trillion U.S. dollars every year, reaching 3% of the global gross domestic product. In particular, the oil, gas and petrochemical industry accounts about one trillion U.S. dollars annually. Significantly, according to a study of the U.S. ExxonMobil, around 40–60% of pipework maintenance costs were due to CUI, in which the costs directly resulting from discontinuation have not even been counted (Wilds 2017). In the oil and gas industry, CUI was assumed to be one of the main reasons for unscheduled plant shutdowns, and it was responsible for more downtime than all other causes. In some extreme cases resulting from CUI, the leakages even led to fires or explosions, which endangered human life. The widely accepted view is that CUI stays more severe in an aging facility than in a relatively new one, and CUI could become a problem if the insulated equipment works for more than five years (Javaherdashti 2014).

By its very nature, CUI remains highly undetectable and unpredictable. The only 100% foolproof way of knowing whether there are CUI problems with equipment is removing the insulating system, which is not the method engineers or end-users willing to adopt since it is the most inconvenient and expensive inspection method. Generally, the removal of insulation for inspection means suspension of production and economic losses. This problematic situation promotes the development of special techniques which can eliminate the requirement to dismantle the insulation system for inspection entirely. These techniques are developed to be non-destructive, called non-destructive examination (NDE) and testing (NDT) techniques. Some of the NDE or NDT techniques can be used to provide comprehensive CUI inspection of insulated systems, such as infrared (I.R.) thermography, profile radiography, computed radiography (C.R.), neutron backscatter, guided wave ultrasonic technology (U.T.), and pulsed eddy current testing (Eltai et al. 2019).

As much more researches related to CUI are focused on detection and anti-corrosion products, not on the corrosion itself. The impact of working conditions inside the insulation and the exposed atmosphere must be considered in CUI research. Generally, the metal and its environment can determine whether corrosion occurs. As the metal system is usually known, while the local environment under jacketing is highly complicated influenced by many factors, studying on the environmental conditions of CUI might be a breakthrough for exploring CUI problem. We propose to provide an idea on the CUI study (shown in Figure 1), which is to combine laboratory experiments with modeling & simulation: focus on the critical factors which affect the environmental conditions of CUI; then design experiments and numerical simulations to gain sufficient and reliable data; finally develop CUI’s prediction through proper data modeling methods. Thus, in this review, we will briefly introduce CUI and the critical factors, not focusing on the development and application of detection techniques and anti-corrosion products but the research of experimental methods and numerical simulations for CUI studies and the CUI prediction methods. We hope readers of this review can gain essential knowledge about CUI, and those who are interested in solving CUI problems can grasp the research status and get inspired by this review.

Figure 1:

Roadmap of the idea from the authors.

2 The essential knowledge of CUI

As shown in Figure 2, the basic structure of an insulated pipe includes: metallic pipe, generally with protective coating; insulation covering the pipe; jacketing to fix and protect the insulation.

Figure 2:

Structure of insulated pipe.

CUI of piping systems can take different forms with different conditions. When moisture penetrates the insulation material and reaches the metal surface, it initially occurs only at the wet positions, causing corrosion at some local areas. Sometimes, it could be general corrosion when a large metal area is soaked or numerous localized corrosion expand and integrate. CUI may also take the forms of pitting corrosion when intrusive water contains salts, and stress corrosion cracking (SCC) may occur when the sensitive material with tensile stress serves in specific solutions. CUI can be galvanic corrosion in some cases since connected dissimilar metals under insulation are exposed to the corrosive solution (Harrison 1977). It is widely accepted that water infiltration from external sources and condensation is the main reason for CUI. There are also some critical factors affecting CUI in insulated systems.

3 Experimental researches of CUI and mass transfer in insulation

CUI is now one of the hottest topics for all researchers from related industries. In order to better understand CUI, laboratory experiments and field tests are both necessary.

The most used experimental design to simulate CUI of piping in the laboratory is the CUI-cell provided in ASTM G189 Standard “Guide for Laboratory Simulation of Corrosion Under Insulation” (ASTM International 2013). This guide covers CUI laboratory simulations, including general and localized attacks on insulated samples cut from pipe sections exposed to corrosive environments (usually at high temperatures). Additionally, it describes the CUI exposure setup (shown in Figure 3), sample preparation, simulation procedures for isothermal or cyclic temperatures or both, and wet/dry conditions, which are parameters need to be monitored during the simulation, and classification of simulation types. Abayarathna et al. (1997) designed the test apparatus in ASTM G189 and indicated the widely cited result to show the comparison of actual plant CUI corrosion rate and laboratory CUI corrosion rate within a temperature range, as shown in Figure 4.

Figure 3:

Schematic of the CUI-cell (ASTM International 2013).

Figure 4:

Comparison of laboratory corrosion data obtained in the open and closed systems with actual plant CUI corrosion rates measurements (open data points shown is for plant CUI) (ASTM International 2013).

This CUI-cell was used in some experiments for insulated piping systems. To overcome the issue that limited information is available for CUI in marine environments, Caines et al. (2015) provided a methodology for developing statistically significant data to evaluate CUI in harsh marine field conditions. Experiments were carried out from two aspects: accelerated laboratory testing with the same apparatus in ASTM G189, 3-years field testing with 52 test cells available for periodic removal from the test site. Williams et al. developed a risk-based remaining life assessment model of CUI rate, and assets under CUI attack were developed with the experiment design. In a similar way, Bai et al. (2017) used the CUI-cell to evaluate cold spray Al-Al₂O₃ coating by testing its corrosion resistance beneath mineral wool insulation under isothermal, thermal cycling and wet/dry conditions. Kane et al. (2008) also did the same experiments to study and evaluate coatings for CUI.

Except for the guide’s method and the methods that directly build a restored structure to simulate CUI, there are other methods to do the CUI experiments. For example, the wire beam electrode (WBE) method, an electrochemically integrated multi-electrode array, was used in combination with noise signature analysis to monitor corrosive species’ penetration for simulated corrosion under insulation conditions. Corrosion of aluminum exposed under insulation materials such as cotton wool, glass wool, rock wool, and tissue paper has been successfully monitored. A good correlation between the galvanic current maps and the corroded surface was also observed (Aung et al. 2006).

Electrochemical-related technology is always an excellent way to study corrosion. Hou et al. (2020) used the electrochemical current noise (ECN) method to investigate corrosion of carbon steel under mineral wool insulation. To identify the type of corrosion, i.e., uniform or localized corrosion, they used recurrence quantification analysis to extract feature variables from ECN signals, which were applied to develop a random forest model.

Another kind of experimental research method for CUI study is to design unique experiments according to the research object. For example, as CUI is related to many factors, even the influence of water drop rate was studied. Eltai et al. (2016) investigated that an insulated mild steel pipe exposed to three seawater drop rates (10, 30, and 50 drops/min) at room temperature was performed. There is no doubt that pipes corroded due to water and ions penetration through the insulation, and the extent of corrosion was varied, which was attributed to the water drop rates. A 10-year life cycle cost analysis focused on the effect of insulation thickness on the life cost of steel pipes with different diameters was conducted by Ertürk (2016). The insulation materials considered were expanded polystyrene, extruded polystyrene, and rock wool. The authors found that the optimal insulation thickness varies with different pipe diameters. Zwaag and Rasmussen (2017) performed cyclic tests of several insulation materials with a dry/wet phase for each cycle. The authors found that closed-cell and semi-closed-cell insulating materials cause substantial surface area to corrode, resulting in higher corrosion rates due to water entrapment on the metal surface.

Protection methods studies were also necessary. Cao et al. (2019) presented a strategy based on cathodic protection (CP) to protect steel piping from corrosion under insulation. There are few experimental studies investigating the CP of steel by wet insulation. In their study, CP was explored to protect insulated steel specimens using sacrificial zinc anodes. Cao et al. discussed the conditions for effectively conferring CP by wet insulation and estimated the throw power protection achieved using electroplated copper in wet insulation using sacrificial zinc anodes with mild steel samples. The efficacy of CP depended on the ionic resistivity and moisture content to achieve effective ionic conductivity through the insulation. When the zinc was in direct physical contact with the low carbon steel, the CP reached the maximum throwing power through the insulating layer when the area ratio of the zinc to the low carbon steel reached a certain value. Cao et al. (2020) also studied the influence of dissolved metal ions from mild steel corrosion and metallic ions leaching from the insulation on the kinetics of CUI. They also explained why steel corrosion rate could be greater when exposed to insulation containing a significant degree of moisture than steel exposed to a solution (with no insulation).

In addition to direct research on CUI, the research on insulation properties and the ingress moisture process, which is one of the most critical points for CUI, is very much needed. However, those kinds of studies were more carried out in many other areas, including building, clothing, and food industries, where porous materials also play important roles.

As shown in Figure 5, this instrument was designed by Fan et al. (2003) in an experimental study of temperature and moisture content distribution within a porous fiber batt sandwiched by thin inner and outer layers of cover fabrics, which were carried out using a new anti-sweat protection hot plate. It was found that most of the changes in temperature distribution occurred within 0.5 h of the test, and the hygroscopicity of the fibers affected the temperature distribution. Moisture content accumulated over time and was found to be higher in the outer regions of the batt than in the inner regions. The accumulation and distribution of water content combined resulted from moisture absorption, condensation, and liquid water movement. These experimental results were needed to understand the mechanism better and provide a verification basis for further theoretical modeling.

Figure 5:

Experimental design on moisture absorption and condensation of fiber insulation materials at low temperature. Reprinted with permission from (Fan et al. 2003). Copyright 2003 Elsevier.

However, the organic nature of cellulose fibers makes the insulation sensitive to high moisture content. Vrána and Gudmundsson (2010) investigated the moisture resistance of cellulose insulation when exposed to subzero environments. The author’s research focused on condensation and freezing in the material and compared to his previous research on asbestos insulation. While the asbestos samples were water-repellent due to resin binders, cellulose is a typical representative of hydrophilic thermal insulation, so contact with condensed water is critical. The results showed a slight change in the water vapor permeability of the samples. They compared experimental data from the survey with a mathematical model that uses actual climate data to simulate the moisture diffusivity of cellulose and the accumulation due to adsorption and freezing.

Peuhkuri et al. (2008) conducted an experimental study to investigate the non-isothermal moisture transport due to temperature gradients on a range of porous light-weight building materials, monitoring moisture fluxes and damp heat conditions generated around and within materials. There were indications that the temperature gradient would drive moisture from the warmer side to the cooler side. All materials used in these experiments, including barely hygroscopic materials (e.g., rock wool) and very hygroscopic materials (e.g., cellulose insulation), showed nearly identical properties.

Ferroukhi et al. (2016) studied the coupled transfer of heat, air, and moisture in multi-story building materials, which was also a typical design in an insulating system. They designed an experimental setup to test the damp-heat behavior of several configurations of multi-layer porous building materials. Heat and moisture transfer in the sample were monitored over time. The adsorption-desorption curves, water vapor permeability, thermal conductivity and specific heat were collected and evaluated.

We can find that these methods are for particular purposes with the aforementioned experimental research methods. In other words, they are one-sided: generally for studying only one or two critical factors of CUI. And except for the CUI-cell, which in the guiding document, other methods are not widely used. The experimental research methods of CUI still need to develop and obtain more useful information from the experiment, such as the environmental conditions to which metals are exposed.

4 Prediction methods of CUI based on field data

Although many researchers and engineers have tried their best in developing inspection and prevention techniques, CUI remains a vast problem that has never been completely solved. Another research direction to address the CUI problem is to predict it, some researchers have proposed CUI prediction methods. In a piping system, the CUI prediction method estimates the lifetime of piping for CUI potential. Two degrees of prediction can be delivered: a conservative prediction, which predicts CUI occurrence earlier than the maximum design life of the equipment; an aggressive forecast, which predicts specifies the maximum life of CUI occurrence (Burhani et al. 2014). However, there is no perfect prediction to solve the CUI problem since both prediction methods have their advantages and risks, especially in terms of system safety and cost. For instance, extreme predictions save budgets but can be very dangerous, leading to higher treatment costs. The importance of the CUI prediction model is to well predict the occurrence and development of corrosion under insulation while ensuring the effectiveness of asset integrity management. Most studies estimate and predict CUI potentials based on system corrosion rates or failure probabilities calculations.

The available corrosion prediction model focuses on climates such as atmosphere, marine, or other harsh environments, and is meant for other types of corrosion: pitting corrosion (Velázquez et al. 2009), atmospheric corrosion (Pintos et al. 2000) and stress corrosion crack (Shi et al. 2015). Besides precious data accumulated from the experimental research of CUI, the data from the field also plays a vital role in studying CUI. Researchers mainly utilize data-driven models such as recurrent neural network and fuzzy logic (Duraisamy et al. 2007). The fuzzy logic method has been used by many researchers in predicting different types of corrosion to alleviate the industrial maintenance system suffered by corrosion.

He et al. (2012) estimated the corrosion rate of L245NB steel in soil by establishing radial basis f unction (RBF) neural network and adaptive neural-fuzzy inference system (ANFIS) models based on corrosion data from modeling emerging corrosion tests. The precision of two methods were compared and the results showed ANFIS was better. Singh and Markeset (2009) proposed a fuzzy logic framework approach to establish the risk-based inspection (RBI) program for piping systems. Wu et al. (2013) developed a model based on fuzzy set theory for risk analysis of corrosion failures for equipment. A fuzzy risk graph was established to determine the risk index based on a fuzzy logic framework. Sa’idi et al. (2014) proposed fuzzy logic system for risk modeling of the process operations in the oil and gas refineries, whose merit was to overcome the uncertainty of the risk-based maintenance (RBM) components and effectively manage failure risk. Bertuccio and Biezma Moraleda (2012) presented the modeling for predicting the probability and severity of the consequences of corrosion in natural gas pipelines by combining fuzzy logic theory and expert judgment. Javaherdashti et al. (2012) developed a composite fuzzy function model to predict microbiologically-influenced corrosion (MIC) of duplex stainless steel in a biotic environment containing corrosion-related bacteria iron reducing bacteria and a control, abiotic synthetic seawater environment. De Moura et al. (2008) focused on using fuzzy logic methods in damage prediction as applied to flat structures under corrosion conditions, in which a structural health monitoring (SHM) approach was applied. Hodhod and Ahmed (2014) developed a fuzzy-based model to predict oil pipeline failures due to mechanical, operational, corrosion, third-party, and natural hazards using historical data of pipeline accidents. Kleiner et al. (2006) established the deterioration modeling of buried infrastructure assets by using a fuzzy rule-based, non-homogeneous Markov process. They described how to translate the fuzzy condition rating of the asset into a possibility of failure, which was combined with the fuzzy failure consequences to obtain fuzzy risk of failure throughout the life of the pipe. Jamshidi et al. (2013) developed a model that combined fuzzy logic and relative risk score (RRS) methodology to evaluate the uncertainty involved in the problem of pipeline risk assessment. Biezma et al. (2018) combined fuzzy logic with in-situ inspection data to predict the external corrosion rate of buried pipelines. The pH, resistivity, redox potential, moisture content, sulfate content, and chloride content were used as the inputs and the soil corrosivity level was used as the output of this fuzzy logic model. Then, the corrosion rate of buried pipelines can be calculated via soil corrosivity level.

Mokhtar and Ismail (2011) presented a fuzzy-based model with operating temperature and type of environment as inputs and corrosion rate as output to predict CUI of carbon steel based on the API 581 data. They used the failure function, which indicated a material’s ability to resist failure, based on the relationship between the material resistance and the applied stress to do prediction research of CUI. Besides, the Mokhtar and Ismail (2011), which calculated the failure probability and reliability index, was also applied since the data for CUI cases presented in the API 581 is very limited for both stainless steels and carbon steels (Khan et al. 2015). Khan et al. (2016) also constructed an adaptive neural-fuzzy inference system model, which is one of the techniques of fuzzy-based methods, learning information by the system from a data set. The accuracy of this model for predicting the corrosion rate of carbon steel was evaluated against the given API 581 CUI corrosion rates. Later, Burhani et al. (2018) successfully developed a new CUI prediction model for piping using two stages of artificial neural network (ANN). The final stage of the ANN model has synthetic inputs from field and experiments, final output with mass loss, and a hidden layer consisting of 9 nodes, resulting in the fitness of the objective function by validating the sigmoidal hyperbolic tangent function and the back-propagation method. Based on the obtained results, it was proposed to divide the artificial neural network into two stages with limited data input for prediction and modeling. Furthermore, since ANNs were good predictors for corrosion modeling, it was recommended to use them for other types of corrosion to improve predictions. Based on their fuzzy logic-based method, Mohsin et al. (2019) further developed CUI prediction model for piping systems with 3D surfaces (as Figure 6 shows). In this prediction model, they chose five critical factors: insulation condition, insulation type, environment, pipe complexity, and operating temperature of CUI, as input parameters and CUI corrosion rate as the output parameter. All of the above researches done by Mokhtar et al. are committed to providing quantitative approaches for engineers and being ascertained as a stunning tool for risk-based inspection (RBI).

Figure 6:

3D surface to present relationship of CUI corrosion rate with temperature and type of environment. Reprinted with permission from (Mohsin et al. 2019). Copyright 2019 Elsevier.

Besides the fuzzy logic methods, other prediction methods based on data analysis were studied. Erickson et al. (2010) utilised Weibull statistical analysis to predict the total number of CUI susceptible locations that require repair each year for an oil field on the North Slope of Alaska. Moon et al. (2010) attempted several prediction methods to predict CUI based on actual corrosion case data from petrochemical plants in Japan. They expected the corrosion occurrence through the database modeling method (Tateno et al. 2010) and continued using k-Nearest Neighbour (Yahiro et al. 2011), which calculated the distance between a demanded point and their historical data and increased the accuracy of the estimation. In 2012, Tateno et al. (2012) used the information gain reliability (IGR) method consisting of more than two thousand cases from 20 actual plants in Japan to estimate the corrosion rate of CUI by setting proper thresholds, which were selected from the correlation between the predicted corrosion rates and actual corrosion rates. Furthermore, Tateno and Khaled (2016) also estimated corrosion rates by using self-organizing map (SOM) and gave conclusions about the correlation coefficient, the estimation results and SOM group numbers. We can look forward to their future CUI prediction work and plant-demonstration experiments mentioned at the end of their paper.

A building information modeling (BIM) based approach for predicting CUI was proposed by Tsai et al. (2019). They utilized passive radio frequency identification (RFID) sensors and smart sensing technologies to collect field data and integrate them into the BIM system. A homogeneous corrosion model was also adapted from the corrosion theories to leverage sensor data and BIM elements’ properties. The collected data served as inputs to calculate the corrosion rate, which could be helpful in corrosion management under insulation.

These available prediction methods have in common that they are based solely on each local industrial or site data. The industrial data can vary due to plant location, environment, and many other factors contributing to model variance and reliability. Generally, data from well-controlled laboratory experiments and accomplished models would be more regular and reliable. Experiments can effectively provide data of CUI-related factors, such as temperature and ambient conditions, insulation type, operating temperature, etc. A well-established and accomplished model of an insulated system can produce data even more rapidly than experiments as well as save time and money.

Therefore, it would be great if these analytical models or methods could be supported by data from laboratory tests or realized numerical simulations.

5 Numerical simulation models of heat and moisture transfer

As computer science develops rapidly, numerical models of heat and moisture transfer in porous materials advance and numerical investigations increase in the literature. These numerical simulations can help establish validated models for CUI by calculating, which describes the specific moisture transfer processes and provides environmental conditions of CUI for further corrosion research. Many types of research focused on numerical simulation about heat and moisture transfer in porous materials, particularly in building engineering materials, while there were few studies about CUI. However, these models of heat and moisture transfer in porous building materials can be good references for establishing models of CUI since they have a very similar objective to describe heat and moisture transfer processes in specific materials. The formation of environment for CUI is mainly two processes: liquid transfer and phase change in an insulated system.

Liquid transfer in an insulated system starts when water is introduced from the outside environment and ends when all liquid is at rest. The fluid flow in porous media has very important applications in many engineering fields, such as underground water resources, geological studies, filtration and purification processes, and petroleum industries (Seyf and Rassoulinejad-Mousavi 2011). In many practical cases, porous materials, including insulating materials, are anisotropic in their mechanical and thermal properties, often the result of preferential orientation or asymmetric geometry of particles or fibers. Fluid flow in anisotropic porous media has attracted the interest of many researchers over the past few decades. In the seminal work of Castinel and Combarnous (1974) and Epherre (1975), it was shown that anisotropy of mechanical and thermal properties affects the marginal stability condition and the preferred width of the convective cell.

Lewis (1921), Richards (1931), Philip and De Vries (1957) and Luikov (1964) elaborated phenomenological models to characterize transport in unsaturated porous media. In the field of building, the first technique developed in the 1980s to assess the moisture content of building materials was the famous Glaser method. More complete models were developed by Pedersen (1992) and Künzel and Kiessl (1996), which have taken into account the diffusive transport of liquids and vapors.

Qin et al. (2006) proposed a dynamic model to evaluate the transient heat and moisture transfer behavior in porous building materials. The transfer of heat and moisture was considered and their interactions were also modeled. The Laplace transform was performed on the coupled system, and then the transfer function method was introduced to solve the system of equations. Leskovšek and Medved (2011) proposed a transient heat and mass transfer model, in which the sorption and condensation processes were included. The parametric study model considered different temperature variations, thicknesses, and moisture masses of the insulating matrix. Choudhary et al. (2004) presented a mathematical model of moist transport, including condensation within the insulation in the presence of the wicking fabric. The model was based on volume-averaged equations for unsteady transport of heat, liquid, and water vapor in porous media. The measurement of full sets of heat and moisture transfer and storage parameters of selected thermal insulation materials independent of moisture content were presented by Jerman and Černý (2012). Expanded polystyrene and hydrophobic mineral wool under where CUI prone occurs were selected as reference materials to validate their approach.

In a CUI environment, except for the problem of heat and mass transfer in porous media, the phase change process should be considered. It is widely accepted that theoretical modeling of coupled heat-moisture transfer with phase change in fibrous insulation started with Henry’s work in 1948 (Henry and Sh 1948), but little research has been done on the coupled heat and moisture transfer with phase change until 1980s.

The transient model of condensation in porous media by Fan et al. (2000) was the first to consider the effect of condensate on effective thermal conductivity and radiative heat transfer. Fan and Wen (2002) extended the previous transient model to account for the movement and evaporation of condensates, which was presumed to be the main reason for the large reduction in adiabaticity observed experimentally (Farnworth 1986). Later, Fan et al. (2004) provided an improved model for coupled heat and moisture transfer with phase change and mobile condensates in fibrous insulation. The moisture movement due to the partial water vapor pressure, supersaturated conditions in the condensation region, the dynamic moisture absorption of fibrous materials, and the mobile of liquid condensates were taken into consideration by this model.

Numerical simulations and modelings of heat and moisture transport are popular approaches in calculating environmental conditions in buildings. The application of transient coupled heat air and moisture (HAM) models allowed for a more efficient estimation and better understanding of the impact from moisture released to indoor spaces and moisture exchanged between outdoor and indoor environments (including wind-driven rainwater loads) (Tariku et al. 2007; Tariku and Kumaran 2020). This was possibly due to the fact that the HAM model considered the effect of moisture during the heat transfer through materials. Usually energy simulation models would ignore the moisture effect when conducting the thermal analysis (Mendes et al. 2003).

Talukdar et al. (2007) explored heat and moisture transfer in porous building materials from two aspects. The experimental one provided data for two hygroscopic building materials (spruce plywood and cellulose insulation) exposed to 1-D and transient boundary conditions. A transient moisture transfer (TMT) equipment was built and utilized to collect the experimental data (shown in Figure 7). The simulation part (Talukdar et al. 2008) presented the experimental results of the TMT facility: the distributions of temperature, moisture accumulation, and relative humidity within the two materials under different and repeated changes in air humidity and at additional airflow Reynolds numbers. Comparing experimental data with numerical data, analytical solutions, and numerical sensitivity studies would increase confidence in the experimental dataset.

Figure 7:

(a) Schematic of the transient moisture transfer facility and (b) test section. Reprinted with permission from (Talukdar et al. 2007). Copyright 2007 Elsevier.

Zhang et al. (2015) proposed a semi-conjugate approach computational fluid dynamics (CFD) – HAM, which coupled the CFD model with simultaneous heat and moisture transfer in hygroscopic material. A dynamic mathematical model to simulate the coupled heat and moisture transfer through multi-layer porous building materials was proposed by Qin et al. (2009). In their work, vapor content and temperature were the main driving potentials, and the finite-difference approach did the discretization of the governing equations. Ferroukhi et al. (2016) developed a macro model that combined diffusion, convection, and conduction to predict the hygrothermal behavior of heat, air, and moisture transfer in multi-layer porous building materials’ effects on architectural elements.

Ferroukhi et al. (2017) studied a new method for predicting the heat and moisture transfer in buildings that combined two simulation tools: COMSOL Multiphysics© and TRNSYS. The COMSOL Multiphysics© was used to model heat, air, and moisture transfer in multi-layer porous walls, and the TRNSYS was used to simulate the hygrothermal behavior of the building. The combined software application was called the co-simulation approach (shown in Figure 8), dynamically solved the mass and energy conservation equations of the two physical models. The method presented in this research can also be used in the research of CUI for combining the local and global processes.

Figure 8:

Diagram of the co-simulation process. Reprinted with permission from (Ferroukhi et al. 2017). © Springer-Verlag Berlin Heidelberg 2016.

Though there are cases about many numerical simulations of environmental conditions (heat, air, and moisture) in insulation materials or similar porous media, the numerical simulations are rarely used for CUI research. A specific metal (with or without coating) exposed in a specific environment determines CUI in this insulated system. The numerical simulation methods can be applied to describe the environmental conditions for an insulated system as they did in other systems (such as building systems) through calculation. Once we know the relationship between the environmental conditions and known metals, which experiments can obtain, we can predict CUI as the environmental conditions were calculated.

6 Concluding remarks

This review gives essential knowledge of corrosion under insulation and summarizes the related researches on it. The recognized mechanism and critical factors of CUI are introduced. Some experimental research methods to study CUI and evaluate corrosion rate are presented.

CUI prediction methods have evolved significantly in recent years. However, most of the models are mathematically extracted from actual field data, whose reliability depends on the quantity and quality of the data. The cruel reality for CUI prediction is that not all industrial enterprises would systematically collect and share CUI data publicly. This situation has become a factor restricting the development of the prediction models of CUI on the basis of historical data.

On the other hand, lab experiments in terms of standardized or customized test methods can deepen our understanding of the local corrosion environment and the corrosion dynamics under insulation. Meanwhile, modeling and numerical simulation have been undertaken to study the heat and moisture transfer process in porous materials, especially in building engineering but not much yet in CUI research, demonstrating the feasibility and effectiveness of combining modeling & simulation with experiments. The authors believe that the combination of experiments with numerical simulation is an excellent way to produce a kinetic model for CUI prediction.

More CUI cases may still occur in aging facilities in the foreseeable future as CUI remains an unsolved problem. Not only detection and protection technologies but also prediction and early warning methodologies call for further development to mitigate CUI. The synergy of kinetic model and model from data science can provide a more comprehensive and reliable prediction for CUI. With on-site condition monitoring technology, the comprehensive model can support the on-condition maintenance of the insulated system.