Conjugate Heat Transfer Benchmarking of Refinery Tank Insulation in Coastal Outdoor Exposure

Tank farms are essential to refinery continuity because production, blending, and shipment rarely proceed at the same pace. Yet the same tanks also create a persistent thermal and integrity burden. A heated storage tank behaves as a large exposed heat-transfer surface, and every avoidable watt lost through the shell must be replaced by fuel, steam, or electricity. That replacement duty is not only a cost. It also sustains elevated shell temperatures, increases the consequence of insulation damage, and can leave local hot surfaces that matter during inspection rounds and routine operator access.

For above-ground tanks, the heat path is conceptually simple and operationally difficult. Heat conducts through the steel wall and insulation, then leaves the outer surface through convection and radiation. Those external terms do not remain fixed. Wind, solar loading, and night-sky exchange change through the day, and even modest differences in boundary representation can alter predicted storage heat load and the apparent benefit of insulation [1, 2]. In exposed coastal service, this variation is not a rare disturbance. It is part of the normal operating envelope.

This is why refinery insulation cannot be judged by nominal conductivity alone. Porous insulation performance shifts with moisture, temperature, density, airflow, and aging, so the conductivity used in design may not be the conductivity sustained in service. Reviews of insulation behaviour consistently identify moisture and temperature as dominant drivers of thermal drift, with additional variation introduced by environmental exposure over time [3, 4]. In practice, the signal is often indirect. A system may appear visually intact while the plant records higher heat duty, less stable product temperature, or both.

Moisture ingress makes the problem more serious because it couples heat loss to corrosion risk. A failed seam, damaged cladding lap, or degraded sealant can admit rain, humid air, or condensate into the insulation system. Once moisture is retained, the steel surface may remain wet for extended periods, which supports corrosion under insulation and reduces the thermal resistance that the insulation was supposed to provide. Experimental work on water transport and drying in insulated systems shows that once water bypasses the weather barrier, wetness persistence is governed by coupled evaporation and adsorption processes rather than by simple drainage alone [5, 6]. A damaged cladding seam left untreated through one wet season is a concrete example: the same defect can increase heat leakage, prolong time of wetness, and hide corrosion activity beneath an apparently minor exterior flaw.

Material choice therefore has direct engineering and safety implications. Aerogel composite blankets are attractive where high resistance per unit thickness is needed, and reported conductivity ranges place them among the strongest thermal performers in this class [7]. Polyurethane foam remains appealing because it is lightweight and widely deployable, but its effective conductivity can change with temperature, humidity, and aging associated with gas diffusion in closed cells [8]. Cellular glass occupies a different position. It is commonly selected where moisture tolerance is central, because its closed-cell structure limits water ingress even though its conductivity is typically higher than that of aerogel and new polyurethane systems [9]. These materials are not interchangeable in refinery service, and the preferred option depends on which operational constraint dominates, minimum heat loss, thickness limitation, moisture tolerance, long-term stability, or some combination of these [10, 11].

The Al-Zawiya refinery context helps explain why this comparison matters, but it should be read carefully. Outdoor tanks there operate within a coastal-arid exposure envelope in which ambient temperature can range from roughly 0 ℃ to 48 ℃ and relative humidity can approach 85% [12, 13]. Under such conditions, daily boundary forcing can materially affect heat loss, while any breach at joints or penetrations can initiate the moisture pathway that underlies corrosion under insulation [14, 15]. Corrosion under insulation is widely recognised as external corrosion that develops when water penetrates the insulation system and reaches the steel surface [16], and its management remains difficult because deterioration progresses with limited visual evidence until inspection or failure reveals it [17, 18]. Safety concerns extend further because degraded insulation can also produce local hot surfaces that increase contact-burn risk during routine plant work [19]. For that reason, this study uses the Al-Zawiya setting as an operationally relevant exposure context, not as a fully parameterised plant-specific case study with resolved site flow physics [20].

The literature already provides several important foundations, but not the exact comparison needed here. Resistance-network methods and standards-based calculations remain the normal route for thickness selection because they are transparent and fast. General material reviews explain how conductivity varies with moisture, temperature, and aging. Corrosion-under-insulation studies clarify why moisture retention and time of wetness matter for integrity. What is less common is a controlled Computational Fluid Dynamics (CFD)-based, like-for-like comparison that isolates candidate insulation families under the same imposed forcing and reports decision-facing outputs, specifically late-time heat flux and outer-surface temperature, for an externally insulated refinery tank. That is the gap addressed in this paper.

Accordingly, this study develops a controlled numerical benchmark for three refinery-relevant insulation families, aerogel composite blanket, polyurethane foam, and cellular glass, applied to an externally insulated heated storage tank. The model resolves conduction through the steel wall and insulation, while the outdoor environment is imposed through a mixed convection-radiation boundary rather than through an explicitly solved external flow field. Thickness effects are examined at 50, 100, and 150 mm, and a 150 mm three-layer configuration is included to test whether layering offers a thermal advantage under ideal bonded contact. The main outputs are heat flux and outer-surface temperature, and these are then interpreted through a preliminary techno-economic screening layer suitable for early retrofit comparison rather than full plant-wide investment appraisal.

The contribution of the paper is therefore specific. It provides a controlled CFD comparison of three insulation families under identical forcing, quantifies thickness-related diminishing returns across a practical 50 to 150 mm range, tests a refinery-relevant multilayer concept against single-layer alternatives, and translates the thermal outputs into screening indicators that are meaningful for maintenance and integrity planning under the stated assumptions.

The sections that follow are organised accordingly: Section 2 reviews the relevant literature, Section 3 defines the model and boundary conditions, Section 4 presents the thermal and screening results, and Section 5 concludes with the practical implications and limitations of the benchmark.

3.1 Case definition and insulation matrix

This study evaluates heat loss from an externally insulated heated storage tank using a controlled three-dimensional conjugate heat transfer model developed in ANSYS Fluent. The purpose of the model is comparative rather than plant-wide predictive. It resolves heat conduction through the steel shell and insulation system, while the outdoor environment is imposed through a mixed convection-radiation boundary at the exposed outer surface. This formulation allows the effect of insulation material and thickness to be compared under identical forcing, which is the main requirement of the present benchmark.

The tank geometry was reconstructed from the project design basis and represented as a vertical cylindrical tank with a diameter of 6.16 m and a height of 7.38 m. The shell was modelled as carbon steel with a wall thickness of 8 mm. Three insulation families were considered as single-layer systems: aerogel composite blanket, polyurethane foam, and cellular glass. Each material was evaluated at three total thicknesses, 50, 100, and 150 mm. In addition, one multilayer configuration was assessed at a total thickness of 150 mm using three 50 mm layers arranged from the tank wall to the ambient side as follows: aerogel composite blanket, polyurethane foam, and cellular glass. Stating the order explicitly is essential because the temperature gradient and effective resistance distribution depend on the layer sequence when the materials differ substantially in conductivity. The overall model geometry and the applied thermal boundaries are shown in Figure 1, while the implemented wall and insulation build-up is presented in Figure 2. The material properties used in the baseline CFD model are listed in Table 2, and the full insulation configuration matrix is summarised in Table 3.

Figure 1. Geometry and boundary conditions used in ANSYS Fluent Al-Zawiya refinery heated fuel-oil tank

Figure 2. ANSYS model geometry showing the tank wall and insulation layers configuration

Table 2. Material properties used in the Computational Fluid Dynamics (CFD) baseline (constant properties)

Table 3. Insulation configurations examined

The insulation matrix was selected to address two practical design questions. The first concerns the relative performance of the three materials when applied individually at the same thickness. The second concerns whether a mixed-layer assembly provides a thermal advantage over the best single-layer alternative under ideal bonded contact. The 50 mm configuration was retained because moderate thicknesses are often the first realistic option in retrofit settings where nozzles, cladding details, or access constraints limit the available build-up.

3.2 Material assumptions and interface treatment

The material properties used in the CFD model were held constant within each simulation. This assumption was adopted to establish a controlled thermal baseline and to isolate the influence of insulation selection from other factors such as moisture uptake, aging, gas diffusion in polymer foams, or interface degradation. The resulting rankings should therefore be interpreted as rankings under dry conditions with ideal layer contact rather than as full field-service predictions.

Heat transfer is formulated as transient conduction in solids with continuity of temperature and heat flux at material interfaces. In ANSYS Fluent, this is implemented through coupled wall interfaces at the steel-insulation boundary. Interfaces between adjacent layers in the hybrid configuration are treated as perfectly bonded and free of explicit contact resistance. This assumption prevents installation-dependent contact conductance from obscuring the material-family comparison, and it defines the multilayer case as an ideal thermal baseline rather than a worst-case installation scenario. The adopted conductivity magnitudes and the relative ranking of aerogel composite blanket, polyurethane foam, and cellular glass are consistent with reported ranges for these material classes [51-53].

3.3 Governing formulation and boundary conditions

Heat transfer within the tank wall and insulation system was modelled as transient conduction in solid domains. The internal hot-side condition was imposed directly at the inner steel wall, while the external boundary was represented through combined convection and radiation. The outward surface heat flux can be written as

$q^{\prime \prime}=h_{\text {conv }}\left(T_s-T_{\infty}\right)+\varepsilon \sigma\left(T_s^4-T_{\text {rad }}^4\right)$     (1)

where, $q^{\prime \prime}$ is the outward heat flux per unit area, $h_{\text {conv }}$ is the external convective heat-transfer coefficient, $T_s$ is the outer surface temperature, $T_{\infty}$ is the ambient air temperature, $\varepsilon$ is the surface emissivity, $\sigma$ is the Stefan-Boltzmann constant, and $T_{\mathrm{rad}}$ is the effective radiative surrounding temperature.

Table 4. Boundary-condition set used in the Computational Fluid Dynamics (CFD) baseline

Table 5. Forcing schedule used in the transient simulations

The internal wall temperature was fixed at 90 ℃ to represent hot-service operation. At the external insulation surface, convection was represented by a constant heat-transfer coefficient of 5 W m⁻² K⁻¹, and surface emissivity was set to 0.85. The radiative surrounding temperature was taken equal to the ambient reference temperature. The baseline ambient temperature was 25 ℃. For clarity and reproducibility, the complete boundary-condition set is reported in Table 4, and the forcing schedule used in the documented transient simulations is stated explicitly in Table 5. The representative external film coefficient was selected in a manner consistent with standard cylinder heat-transfer correlations used to translate outdoor exposure into an effective boundary condition when the airflow domain is not solved explicitly [54].

The transient simulation window was set to 24 h, corresponding to 86,400 s. As documented in the uploaded files, the ambient condition remained at 25 ℃ over the full 24 h period. This point is stated directly here so that the paper no longer implies an hourly forcing profile that is not numerically reported. All insulation cases were therefore compared under the same constant thermal forcing over the 24 h transient window, allowing late-time heat flux and surface temperature to be reported on a consistent basis.

3.4 Numerical setup, mesh strategy, and convergence control

All simulations were executed in ANSYS Fluent using a three-dimensional heat-transfer formulation in which only the energy equation is solved within the solid regions, and the environment is imposed through the mixed convection–radiation boundary. Spatial discretisation uses second-order schemes for the energy equation. Time integration uses an implicit transient scheme.

Transient simulations were marched to 86,400 s (24 h) to resolve the conductive response under diurnal forcing and to confirm stabilisation of the reported heat-flux and surface-temperature metrics. Time-step selection was chosen to be materially smaller than the forcing interval (1 h). A sensitivity check was performed to confirm that reducing the time step further does not change the monitored outputs beyond the mesh-study tolerance.

The mesh was designed to resolve the dominant through-thickness gradients in the steel wall and insulation layers. Refinement was concentrated near the inner wall, the steel-insulation interface, the interfaces between insulation layers in the hybrid case, and the exposed outer surface where heat exchange with the environment occurs. A conformal mesh was maintained across all bonded interfaces [55, 56]. This arrangement is appropriate for layered conduction problems because it minimises numerical smearing and captures the thermal drop across each material with greater fidelity. The local refinement pattern adopted for the insulation model is shown in Figure 3.

Figure 3. Mesh refinement using ANSYS for the insulator

Convergence at each time step is assessed using a dual criterion: residual reduction for the energy equation and stability of engineering monitors. The monitors include area-averaged outer heat flux, maximum and average outer surface temperature, and layer-by-layer temperature drops. This matters because the decision variables are integrated quantities, and residuals alone can be misleading in conduction-dominated simulations [57].

3.5 Verification and reproducibility controls

To strengthen reproducibility, three checks were used: analytical consistency, mesh independence, and time-step independence. The analytical check compared the CFD prediction with a cylindrical thermal-resistance calculation under the same boundary assumptions. The deviation was evaluated using

$\operatorname{Deviation}(\%)=\frac{\left|q_{\mathrm{CFD}}^{\prime \prime}-q_{\text {analytical }}^{\prime \prime}\right|}{q_{\text {analytical }}^{\prime \prime}} \times 100$   (2)

A mesh-refinement study was then performed using coarse, medium, and fine grids. Acceptance was based on monotonic convergence and a change of less than 2% in the main reported outputs between successive refinement levels. A similar criterion was applied in the time-step study, where three time increments smaller than the reporting interval were compared. The grid-refinement and verification workflow adopted in the study is summarised in Figure 4, while the resulting numerical checks are consolidated in Table 6. This verification structure follows common practice in discretisation-uncertainty and grid-convergence reporting for CFD-based thermal studies.

Figure 4. Mesh refinement and grid independence workflow for the tank insulation model

Table 6. Verification summary for analytical consistency, mesh independence, and time-step independence

The verification results indicate close agreement between the analytical resistance model and the CFD solution, with a heat-flux deviation of 1.02%, which remains within the adopted acceptance band. The mesh study shows stable monotonic convergence, while the time-step study confirms that the reported 24 h thermal outputs are insensitive to further temporal refinement within the selected range. On this basis, the medium mesh and the 600 s production time step were retained for the full simulation matrix because they provided stable predictions at lower computational cost than the finest settings.

3.6 Output extraction and economic framing

All cases were post-processed using the same reporting convention. The inner reference corresponds to the imposed hot-side wall temperature. For the uninsulated case, the outer wall corresponds directly to the exposed steel surface. For insulated cases, temperatures were also extracted at the steel-insulation interface and, in the hybrid case, at each internal layer boundary. This layer-by-layer reporting is important because it shows how the thermal gradient is redistributed when multiple materials are combined.

Three primary outputs were extracted from the simulations: local and area-averaged outer heat flux, local and area-averaged outer-surface temperature, and total heat loss obtained by integrating heat flux over the external tank surface. These quantities were selected because they connect directly to the engineering questions of the study. Heat flux represents the continuing thermal penalty of hot storage. Outer-surface temperature relates to surface safety and the practical thermal behaviour of the insulation system. The comparison at 24 h was used as the main reporting point because it reflects the late-time condition under identical forcing for all cases.

The economic interpretation of the results was kept at screening level. Present-worth notation was used to maintain a consistent economic basis:

$P V=\frac{C_t}{(1+r)^t}$   (3)

$L C C=C_0+\sum_{t=1}^n \frac{C_t}{(1+r)^t}$   (4)

where, $P V$ is present value, $C_t$ is the cost or saving at time $t, r$ is the discount rate, $C_0$ is the initial investment, and $n$ is the evaluation horizon [58]. In the present study, however, the economic comparison is intentionally limited to a preliminary screening perspective derived from the CFD thermal outputs. It is not presented as a full refinery-wide techno-economic appraisal, which would require site-specific installation cost, labor, maintenance frequency, degradation history, and downtime exposure.

This work developed a controlled CFD benchmark for comparing three refinery-relevant insulation families, aerogel composite blanket, polyurethane foam, and cellular glass, for an externally insulated heated storage tank under a consistent mixed convection-radiation boundary. The uninsulated reference confirmed the severity of baseline heat loss, with a 24 h external heat flux of 783.77 W/m². Once insulation was introduced, heat flux dropped sharply and the ranking by thermal performance remained stable across thickness. Aerogel composite blanket gave the lowest heat flux at every thickness, polyurethane foam remained intermediate, and cellular glass was consistently higher.

A second practical finding was the presence of a clear thickness knee. Increasing thickness from 50 to 100 mm produced a strong additional reduction in heat flux and outer-surface temperature. Increasing thickness from 100 to 150 mm remained beneficial, but the marginal gain weakened because the external convection-radiation boundary increasingly limited further improvement. Under the stated assumptions, this makes the 100 mm range a credible engineering compromise for many retrofit settings, especially where thermal benefit, constructability, and material use must be balanced together.

The multilayer 150 mm configuration provided an important cautionary result. Although it combined three insulation families, its final heat flux, 25.307 W/m², was essentially equal to the 50 mm polyurethane foam case and clearly above the 150 mm single-layer aerogel and polyurethane cases. Under the stated assumptions, the hybrid assembly did not produce a purely thermal advantage, because the inclusion of a higher-conductivity layer reduced the total thermal resistance of the stack. Its justification, therefore, lies in system-level considerations such as constructability, moisture tolerance, or exposure management, not in an assumed thermal synergy.

When interpreted through a preliminary screening metric, polyurethane foam emerged as the most attractive performance-to-cost compromise, aerogel composite blanket remained the thermal best with a large cost premium, and cellular glass remained a defensible choice where durability and moisture resistance dominate the selection envelope. These conclusions should be read as a reproducible thermal baseline for comparison and early retrofit screening, not as a full plant-specific field validation. Future work should extend this baseline by introducing sensitivity to external coefficient variation, radiative conditions, interface resistance, and moisture-degraded conductivity so that ranking stability can be tested under more realistic service variability.

Conceptualization, J.A.J.A. and A.A.S.A.S.; methodology, H.T.M.D.; software, J.A.J.A.; validation, H.T.M.D.; formal analysis, M.F.; investigation, J.A.J.A.; resources, M.F.; data curation, J.A.J.A.; writing—original draft preparation, J.A.J.A. and S.G.H.; writing—review and editing, All authors; visualization, A.A.S.A.S.; supervision, M.F.; project administration, H.T.M.D.; funding acquisition, M.F. and S.G.H. All authors have read and agreed to the published version of the manuscript.