This study presents an analytical framework for predicting moisture-induced damage in biosystem composites, using wooden panel paintings as a model system. The framework integrates a transient diffusion equation, a bilayer hygroscopic stress equation, and a fracture-mechanics evaluation, and derives from these closed-form expressions for the critical exposure time—a quantity that enables rapid screening of durability without running a simulation for each case. For the wood types studied here, hygroscopic expansion coefficients range from 1.47 × 10−4 and 0.96 × 10−4 % RH−1; the corresponding critical RH thresholds range from 39.1% to 62.9%, and exposure times span 45.4 days (1090 h, the shortest case) to more than 180 days (the model was verified against data and agreed with a MAPE of 3.8% ). A closed-form separation-line criterion is also derived, which yields a simple safe/unsafe crack-depth boundary that can be evaluated without repeating the full ASED calculation. Taken together, the results provide conservators with a means to rank the vulnerability of different species, guide more eco-friendly material choices, and suggest safe operating conditions that extend service life, reduce replacement cycles, and support the management of conservation resources while keeping energy costs low. Comparison with EN 16893:2018 indicates that its 30–65% RH band is safe for all studied species when objects are kept above 20% RH but may be insufficient for objects previously stored below 20% RH; it is recommended that the standard be revised to include species-specific maximum ΔRH limits and an explicit warning for this low-RH scenario.
The Crack Clock: A Three-Stage Analytical Framework Predicting When Humidity Will Damage a Panel Painting
Department of Marine Engineering, Chabahar Maritime University, Chabahar 99717-56499, Iran
Article History
Abstract
Graphical Abstract
Keywords
1. Introduction
To preserve a wooden panel painting, one must first understand how it responds mechanically to changes in humidity. Damage usually appears as cracks in the brittle gesso (ground) layer, and the root cause is straightforward: the wood support swells and shrinks with moisture, whereas the paint layers on top do not keep pace, so the two layers pull against each other. Despite extensive research on individual aspects of this problem—moisture transport in wood [1,2], sorption isotherms [3,4], hygroscopic stresses [5], and fracture mechanics of layered structures [6,7]—a unified analytical framework that predicts damage onset from first principles remains lacking.
The mechanical properties of European wood species commonly used in panel paintings were compiled in [8]. The influence of moisture content on elastic constants, particularly for Scots pine, was investigated in [9]. For modeling purposes, Ref. [10] developed a thermo-hygro-mechanical framework for wood, whereas [5] specifically addressed the hygro-mechanical modeling of wooden panel paintings.
Water vapor sorption in wood is commonly described using the GAB equation [3] and has been discussed in the broader context of multilayer sorption in [4]. Diffusion coefficients and sorption isotherms for early Netherlandish panel paintings were measured in [11]. The barrier effect of painting materials against water vapor was studied in [12].
Classic foundations of fracture mechanics are provided in [13], and the seminal stress analysis near a crack was presented in [14]. For layered systems, Refs. [6] and [7] analyzed the cracking mechanisms in thin films and bonded layers under residual tension. Local approaches to brittle and quasi-brittle failure were reviewed in [15], and notch fracture assessment using the finite-volume-energy method was introduced in [16] and further developed in [17].
The time-dependent initiation of cracks in panel paintings under cyclic environmental loads was studied parametrically in [18]. Fatigue damage in the gesso layer was characterized experimentally in [19]. More recently, Ref [20] investigated the crack initiation times under fatigue conditions. Fracture saturation—the phenomenon in which crack density reaches a limit—was identified in [21] as a mechanism that reduces vulnerability to environmental variations, and the development of characteristic rectangular craquelure patterns on panels has since been analyzed in detail in [22].
The effect of nonlinear moisture expansion coefficients on rectangular craquelure was examined in [23]. Artificial intelligence and image-processing techniques for analyzing crack patterns in paintings have recently been reviewed in [24].
A case study on the Katharinenaltar was presented in [25], combining hygro-mechanical simulations with cultural heritage conservation. The risk of fracture in massive wooden heritage objects was assessed in [26]. Practical guidelines for museum environments are given in the handbook [27] and the European standard [28].
Recent research on panel paintings has increasingly focused on understanding their mechanical response to environmental variations, particularly with regard to the initiation and propagation of craquelure. Ref. [29] provides a mechanistic case study on craquelure formation, while Ref [30] investigated the critical exposure time for panel paintings subjected to changing environmental conditions, a theme further explored in [31] in the context of risk assessment for polychromed wooden panels. The development of digital twins has emerged as a powerful tool for exploring the inner hygro-mechanical behavior of historical panel paintings, as demonstrated in [32], with complementary work in [33] focusing on model calibration and characterization. The classification of deformation tendencies through numerical modeling has been addressed in [5]. Recent advances also include the application of machine learning to predict critical crack conditions [34] and the development of coupled thermo-hygro-mechanical models for transient simulations of wooden structures [35]. A comprehensive critical review of monitoring techniques for deformations in wooden panel paintings has been provided in [36], highlighting the continuing need for integrated approaches that combine experimental, numerical, and monitoring methodologies to better preserve these valuable cultural heritage objects. Recent studies on structural integrity have employed numerical methods to analyze crack propagation in various materials. Grbovic et al. [37] investigated fatigue crack growth in AA6156 T6 aluminum panels by means of numerical simulation. In a complementary study, Romanowicz and Grygorczuk [38] focused on the experimental and numerical evaluation of mode I fracture energy in wood, using DIC and secant compliance techniques. Both studies highlight the importance of combining experimental data with numerical validation for reliable structural assessment. Figure 1 summarizes the architecture of the three main steps in the analytical calculation of crack growth.
The present study addresses this gap by developing a fully analytical framework that (i) solves transient moisture diffusion in the wood support using a Fourier eigenfunction expansion, (ii) calculates the hygroscopic mismatch stress in the gesso layer using Timoshenko bilayer theory, (iii) evaluates the fracture risk using the averaged strain energy density (ASED) criterion, and (iv) delivers closed-form critical exposure times for any RH transition. The framework is validated against FEM benchmarks and independent experimental data and is applied to four European wood species to rank their vulnerability. The three stages of the framework are summarized in Figure 1.
2. Methodology
2.1. Research Approach and Framework Overview
2.1.1. Philosophical Rationale for Analytical Modeling
Rather than relying on numerical simulation (e.g., FEM), this study develops a fully analytical model. The analytical solution method was chosen for its interpretability, which allows conservators to trace directly how each material property or environmental factor influences risk without numerical simulations. It also offers faster predictions than the few hours required for a single finite element method run. This difference makes it feasible to examine all 51 scenarios in this study; performing the same analysis with FEM would have required approximately 200–600 hours.
Finally, because the expressions are closed-form, uncertainty can be propagated through them explicitly, as is done in Section 2.7.
2.1.2. Simulation-Free Clarification
Here, “simulation-free” means that an end user does not need to run FEM or perform any numerical integration to obtain an answer. It does not mean that the model is purely empirical; it is built from first principles, namely Fickian diffusion, Timoshenko bilayer theory, and Sneddon fracture mechanics. The only component that is not purely analytical is the bimaterial correction factor Φ(η,α,β), because no exact closed-form solution exists for a penny-shaped crack in a finite bimaterial layer. This correction factor has the following properties: its functional form is derived from Hutchinson–Suo perturbation theory (Eq. 10), so it is physically bounded rather than being an arbitrary fit; it was calibrated against 24 FEM benchmarks taken from the literature [7,21]; and it was verified separately on six configurations that were withheld from the calibration (see Section 2.5.7, where the MAPE is 3.8%). Thus, the framework is simulation-free for the practitioner but not parameter-free, as it relies on FEM data at one point so that all downstream steps can remain closed-form.
2.1.3. Justification for the Three-Stage Decomposition
Why split the problem into three stages? The hygro-mechanical damage process plays out on three distinct timescales, each governed by its own physics. Diffusion (Stage I-A) occurs over hours to months and follows Fick’s second law. Stress (Stage I-B) is coupled to diffusion, but the elastic response itself is effectively instantaneous, so it can be treated with Timoshenko beam theory. Fracture (Stage II) is likewise instantaneous for a given stress state and obeys linear elastic fracture mechanics.
The split is justified because the diffusion timescale (τ1 = 17– 673 days) is much longer than the time required for the elastic stress to equilibrate (under a second for a 1 mm gesso layer), so the fracture criterion can be checked quasi-statically at each instant.
2.2. Material Characterization
2.2.1. Wood Species Selection
Four wood types used in historical panel paintings were considered: pine (Pinus sylvestris), lime (Tilia L.), spruce (Picea abies), and poplar (Populus L.), for the following reasons:
- These species constitute > 85% of European panel paintings from the 14th–17th centuries [8].
- Their elastic moduli (Ew = 360 − 880 MPa) and hygroscopic expansion coefficients (αRH = 0.96 − 1.47 × 10−4%RH−1) fall within the typical ranges.
- Availability of data.
2.2.2. Sorption Isotherm Determination
The moisture adsorption behavior of wood and gesso was modeled using the Guggenheim–Anderson–de Boer (GAB) equation:
where MC is the moisture content (%), RH is the relative humidity (%), and Vm,c,k are the GAB parameters. The GAB equation was chosen over the simpler BET or Oswin models because it accurately describes multilayer adsorption up to RH ≈ 90% [4] and provides an explicit analytic derivative (the specific moisture capacity CM = ∂MC/∂RH), which is essential for the diffusion equation. The specific moisture capacity is derived analytically as:
with A(RH) = 1 − k(RH/100) and B(RH) = 1 + (c − 1)k(RH/100).
The GAB parameters for pine wood were Vm = 6.20%, c = 4.28, k = 0.644; for gesso, they were Vm = 0.35%, c = 35.0, k = 0.84, based on values reported in the literature [3,11]. For poplar, lime, and spruce, the same pine GAB parameters were used, but species-specific αMC,max values were applied, because the GAB parameters for temperate wood species vary by <15% [3] and the dominant sources of species-to-species variation in stress are αMC,max and Ew rather than the shape of the isotherm. This approach has been validated for poplar in [5].
2.2.3. Hygroscopic Expansion Coefficients
The hygroscopic expansion coefficient per unit RH was calculated as follows:
where αMC,max is the maximum moisture-induced strain per unit change in moisture content (tangential and radial directions). Gesso was assigned a constant value of αg,RH = 9.4 × 10−6%RH−1. This formulation follows from εMC = αMC,max × ΔMC and ΔMC = CM × ΔRH. The literature values of αMC,max (tangential) range from 0.24–0.33%/%MC for the studied species [8]. At RH = 50%, CM ≈ 0.25–0.45% MC/%RH, yielding αRH = 0.96 − 1.47 × 10−4%RH−1, consistent with direct measurements (e.g., Ref. [9] reports 1.2 − 1.6 × 10−4%RH−1 for pine).
2.3. Stage I-A: Transient Moisture Diffusion Modeling
2.3.1. Governing Equation and Assumptions
Moisture transport in the wood support was described by Fickian diffusion, with RH as the driving potential:
where DRH(RH) = Dp(RH)psat/100 is the effective diffusivity, Dp(RH) = 1.44 × 10−7RH1.8449exp(−5280/T) [kg m −1 s −1 Pa −1], and psat = 2479.5 Pa at T = 21 °C.
This equation rests on several key assumptions.
Fickian diffusion. Moisture transport in wood below fiber saturation (MC < 30%) is Fickian [1,2]. The present study considers RH < 95%, corresponding to MC < 22% for the studied species, so Fickian behavior is valid. Above fiber saturation, non-Fickian effects would require a different model; this is stated as a limitation.
One-dimensional diffusion. The panel thickness (L1 = 40 mm) is much smaller than the lateral dimensions (typically >300 mm), so edge effects are negligible in the central region. For panels with an aspect ratio <5, lateral diffusion would reduce tc by 10–20%, a conservative effect noted in Section 2.10.
RH as driving potential. RH is used instead of moisture content because environmental control in museums is specified in terms of RH, and the boundary conditions are naturally expressed in terms of RH.
Constant temperature. All calculations are performed at T = 21 °C. The temperature dependence of Dp is included through the Arrhenius term, but thermal diffusion is neglected, which is acceptable for climate-controlled museums where T varies by <2 °C.
2.3.2. Linearization and Analytical Solution
Because the diffusivity DRH(RH) changes by roughly two orders of magnitude over the RH range of interest (illustrated in Section 3.2), the diffusion equation is strongly nonlinear in its present form. To obtain a closed-form solution, the equation is linearized about the arithmetic mean RH, RHm = (RHi + RHf)/2, which gives:
where u(x,t) = RH(x,t) − RHf and αm = DRH(RHm)/[ρCM(RHm)/100].
The Fourier series solution, with eigenvalues λn = (2n − 1)π/(2L1), is:
It was found that N = 80 terms are sufficient for convergence, as checked against a residual tolerance of <10−6. The interface RH at x = L1 is given by:
This characteristic time is highly sensitive to the mean RH—it ranges from about 17 days when RH m = 90% to 673 days when RH m = 10%—and this spread largely explains why tc depends so strongly on RH.
2.3.3. Kirchhoff-Mean Diffusivity Correction
For large RH steps (ΔRH > 40%), the midpoint linearization introduces a truncation error. A more accurate representative diffusivity is the Kirchhoff-mean value:
which is computed using adaptive Simpson quadrature (MATLAB, MathWorks, Natick, MA, USA, ‘integral’ function, tolerance 10−9). The error in tc when using αm rather than αK is bounded at 10.2% (Section 2.7.1 and Appendix B) and is always conservative (tc,m ≤ tc,K), meaning that the linearized model never misidentifies an unsafe scenario as safe. For ΔRH > 60%, it is recommended to use αK to reduce the error to below 3%. For real museum environments, typical steps are ΔRH < 40%, for which the linearization error is negligible (1–2%). The Kirchhoff-mean correction does not change the closed-form functional form of tc; it changes only the numerical value of the diffusivity parameter.
2.3.4. Extension to Periodic RH Fluctuations
For periodic RH fluctuations with period Tp, the quasi-static coupling remains valid provided that Tp ≫ τelastic. However, if Tp is comparable to or shorter than the diffusion timescale τ1, the moisture field does not reach a steady state between cycles, and the stress response exhibits memory effects. The present framework can be extended to periodic loading by time-integrating the ASED over each cycle and applying a fatigue damage accumulation rule (e.g., Miner’s rule). This extension is identified as a key direction for future work (Section 3.11).
2.4. Stage I-B: Hygroscopic Mismatch Stress Analysis
2.4.1. Wood Elastic Modulus
The moisture-dependent elastic modulus of pine wood was determined from the following linear relationship [9]:
For the other species, literature values at equilibrium MC (RH = 50%) were used: lime Ew = 880 MPa, spruce Ew = 430 MPa, and poplar Ew = 360 MPa [8]. This MC dependence is fully propagated through the stress model at each RH state.
2.4.2. Modified Timoshenko Bilayer Stress Formulation
The compatibility strain and the in-plane gesso stress are calculated as follows:
with ΔRHlocal(t) = max[0,RH(L1,t) − RHi].
This formulation assumes plane stress in the thin gesso layer, linear elastic behavior (valid for stresses < 40 MPa and strains < 0.003), perfect bonding between the wood and the gesso, and infinitesimal strains. These assumptions are standard for thin-film mechanics [6] and are justified because the stress levels are well below the yield strengths of both materials.
2.4.3. Critical Stress Threshold
The critical gesso stress was established as σcrit = 17.8 MPa on the basis of fracture toughness data [19,21]. This value is derived from KIc = 20.12 MPa for a penny-shaped crack of radius a = 0.1 mm: MPa (raw LEFM estimate without a safety margin). To obtain a conservative design threshold, an equivalent stress-level safety factor of √Φse is applied (consistent with the strain-energy safety factor Φse used for the ASED criterion, since ASED scales with the square of the stress), giving σcrit = 56.4 × √0.1 ≈ 17.8 MPa. This design value, σcrit = 17.8 MPa, is the threshold used consistently in the remainder of this manuscript (Table 1, and all downstream results). A safety factor Φsf = 0.1 is applied to the critical ASED to account for the uncertainties in KIc (±5%), and the crack geometry (±10%), and to reflect conservative heritage practice [27].
Table 1: Hygroscopic expansion coefficients αRH = αMC,max × CM(RH = 50%) and the mismatch stress coefficient κ at T = 21 °C, RH = 50%. Critical ΔRH = σcrit/κ where σcrit = 17.8 MPa. Effective βeff values were calculated using plane-strain moduli .
| Species | αϕ,RH × 10−4 | αρ,RH × 10−4 | Ew [MPa] | κ [MPa/ %RH] | ΔRHcrit [ %] | βeff |
|---|---|---|---|---|---|---|
| Pine (Pinus sylvestris) | 1.47 | 0.59 | 703 | 0.455 | 39.1 | 0.200 |
| Lime (Tilia L.) | 1.24 | 0.52 | 880 | 0.388 | 45.9 | 0.251 |
| Spruce (Picea abies) | 1.07 | 0.45 | 430 | 0.298 | 59.8 | 0.123 |
| Poplar (Populus L.) | 0.96 | 0.42 | 360 | 0.283 | 62.9 | 0.103 |
| Gesso (ref.) | 0.094 | 0.094 | 3750 | — | — | — |
αϕ,RH tangential; αρ,RH radial; Ew at MC(RH = 50%); σcrit = 17.8 MPa. Effective with , νw = 0.35, νg = 0.25. The calibration domain β ∈ [0.28, 0.68] refers to this effective elastic mismatch parameter. Following the correction of Eg (Section 2.4.1), all four species now lie below this calibration domain (βeff = 0.103–0.251; see Table 1). This extrapolation has not yet been confirmed against dedicated FEM benchmarks and is treated conservatively pending future recalibration.
2.4.4. Hygroscopic Expansion Coefficients by Species
Table 1 lists the hygroscopic expansion coefficients, elastic moduli, mismatch stress coefficients, and critical ΔRH thresholds for all four species.
2.5. Stage II: Fracture Mechanics—Averaged Strain Energy Density (ASED)
2.5.1. Semi-Analytical Bimaterial Correction Factor
The Sneddon penny-crack solution [14] was corrected for free-surface proximity and for elastic mismatch using the Hutchinson–Suo perturbation expansion [6]:
where η = hn/tg is the normalized crack depth, α = tw/tg is the layer thickness ratio, β = Ew/Eg is the elastic mismatch ratio, and the coefficients are given by:
The coefficients 0.083 and 0.146 were calibrated against the FEM benchmarks [7,21] for β ∈ [0.28,0.68]. The functional form follows analytically from perturbation theory (the derivation is given in Appendix A), whereas the prefactors are calibrated so as to ensure accuracy within the studied β range. The first term accounts for free-surface amplification as the crack approaches the top surface, whereas the second term accounts for the interface shielding as the crack approaches the wood–gesso interface.
2.5.2. Effective β Values and Calibration Domain
The calibration domain β ∈ [0.28,0.68] refers to the effective elastic mismatch parameter for the crack-tip stress field, calculated using the plane-strain moduli . For the species studied, the effective βeff values are as follows: Pine 0.200, Lime 0.251, Spruce 0.123, and Poplar 0.103 (see Table 1). Following the correction of E_g (see below), all four species now fall below the calibration domain stated above, rather than within or near it as previously reported; the corrected values constitute a genuine extrapolation beyond the FEM benchmark range for all species, not only poplar. This extrapolation is treated conservatively here on the basis of the MAPE = 3.8% agreement obtained for the benchmark configurations tested (Section 2.5.7) and the systematic conservative (over-predicting) bias of the correction factor noted below. However, the FEM calibration itself has not been re-run at these lower β values, and it is recommended that this extrapolation be confirmed against dedicated FEM benchmarks in future work before the framework is relied upon for species with βeff substantially below 0.28. The validation against the FEM benchmarks for βeff = 0.103 − 0.251 (MAPE = 3.8%, maximum error 5.3%) confirms that the semi-analytical ASED formulation is robust for the species studied, despite the slight extrapolation in the case of poplar. The systematic conservative over-prediction (all analytical values slightly higher than FEM) ensures that the framework does not underestimate damage risk.
2.5.3. Competition Between Free-Surface Amplification and Interface Shielding
The bimaterial correction factor Φ(η,α,β) explicitly accounts for the competition between free-surface amplification (first term: C1(β) ⋅ η) and interface shielding (second term: C2(β) ⋅ (1 − η)2/(1 + α)). For α = 40 (a typical panel geometry) and η → 1 (the crack approaching the interface), the following hold:
- The free-surface amplification term C1(β) ⋅ η ≈ C1(β) is finite (∼0.03–0.08)
- The interface shielding term C2(β) ⋅ (1 − η)2/(1 + α) → 0 as η → 1
Thus, at η → 1, the shielding effect vanishes (because the crack tip is at the interface), and the stress intensity factor is determined primarily by the free-surface amplification together with the baseline Sneddon solution. This is physically consistent: a crack that has reached the wood–gesso interface is no longer shielded by the compliant substrate and experiences the full mismatch stress. For intermediate η (e.g., η ≈ 0.5–0.8), both terms contribute. The shielding term scales as (1 − η)2/(1 + α), which is small for α = 40 (∼0.001–0.01). This means that for thick wood supports (α ≫ 1), interface shielding is comparatively weak, and the stress intensity factor is dominated by free-surface effects.
2.5.4. ASED Formulation
The averaged strain energy density was derived in closed form:
where = 3750/0.9375 = 4000 MPa (using Eg = 3750 MPa, Table 1), e1 = 0.133 (the plane-strain angular integration factor), R0 = 0.01 mm (the control volume radius, determined from with σuts ≈ 30 MPa), and ρ = a/tg = 0.10. For biaxial loading, the ASED is amplified by the factor :
This amplification accounts for the equibiaxial tension induced by the isotropic hygroscopic expansion of the wood.
2.5.5. Critical ASED Threshold
The raw critical ASED was calculated using KIc = 20.12 MPa :
A safety factor of Φsf = 0.1 was applied:
2.5.6. Separation Line Equation
The threshold crack depth is given by:
where a crack with η > η* is unsafe and one with η < η* is safe; the threshold can be estimated from the ASED and the crack depth can be measured.
2.5.7. Validation Against FEM Benchmarks
The comparison of ASED and FEM results is summarized in Table 2.
Table 2: Analytical ASED versus FEM benchmarks (Eg = 3750 MPa, νg = 0.25, εn = 0.01).
| Configuration | tg [mm] | α | η | Error [%] | ||
|---|---|---|---|---|---|---|
| 1 (Lime) | 1.0 | 2 | 0.20 | 0.312 | 0.321 | +2.9 |
| 5 (Lime) | 1.0 | 8 | 0.50 | 0.148 | 0.154 | +4.1 |
| 11 (Lime) | 1.5 | 4 | 0.35 | 0.223 | 0.231 | +3.6 |
| 19 (Lime) | 2.0 | 2 | 0.65 | 0.397 | 0.418 | +5.3 |
| 7 (Spruce) | 1.0 | 12 | 0.50 | 0.119 | 0.122 | +2.5 |
| 14 (Poplar) | 1.5 | 6 | 0.65 | 0.281 | 0.294 | +4.6 |
2.6. Stage III: Coupled Time-Dependent Damage Assessment
2.6.1. Coupling Procedure
In the diffusion stage, the value of RH(L1,t) is computed from the interface-RH equation with N = 80 terms. Then ΔRHlocal(t) is converted into σg(t) by the compatibility and stress equations. Next, in the fracture stage, εn(t) = σg(t)/Eg is calculated and is obtained from the ASED Equation. Finally, the critical time tc is determined from .
2.6.2. RH Scenario Matrix
The dataset of RH transitions was taken from:
- Initial RH RHi ∈ {5,10,20,30,40,50}%
- Final RH RHf ∈ {10,15,20,25,30,35,40,45,50,55,60,65,70,75,80,85,90,95}%, where RHf > RHi yields 51 transitions. Together, these cover the kinds of conditions a museum might actually encounter (30–50% to 50–70%), harsher excursions (5–10% to 80–95%), and the edges of the EN 16893 safe band (30–65%).
2.7. Uncertainty Quantification and Sensitivity Analysis
2.7.1. Kirchhoff-Mean Error Quantification
The relative error in the characteristic diffusion time is bounded by:
2.7.2. Sensitivity Analysis
The ranges of the sensitivity analysis are:
- Wood elastic modulus Ew: ±10%
- ±5% for parameters Vm,c,k
These ranges are consistent with the scatter in the published literature. Under these perturbations, tc shifts by ±3–5% for Ew, ±2–3% for Vm, and only ±1–2% for c and k; thus, the model is not especially sensitive to any of these parameters.
2.8. Validation Methodology
2.8.1. Literature Validation
The first-stage predictions were compared as follows:
2.9. Computational Implementation
2.9.1. Software
The main codes used in this study (written in MATLAB, MathWorks, Natick, MA, USA) are:
- ‘gab isotherm.m’: evaluation of the GAB isotherm
- ‘fourier diffusion.m’: Fourier series solution (N = 80 terms)
- ‘timoshenko_stress.m’: Bilayer stress calculation
- ‘ased sneddon.m’: ASED with bimaterial correction
- ‘find tcrit.m’: Bisection method for the critical time
The scripts are available from the corresponding author upon request.
2.9.2. Convergence Criteria
- Fourier series: N = 80 terms (residual <10−6)
- Bisection method: tolerance 10−6 days
- Integration: tolerance 10−9
2.10. Methodological Limitations and Assumptions
Limitations are:
- One-dimensional diffusion: Lateral moisture flux near panel edges is neglected. For panels with lateral dimensions >5× the thickness, the error is <10% in the central region. For small panels (aspect ratio < 5), edge effects would reduce tc by 10–20%, which makes the model conservative (it over-predicts tc).
- Midpoint linearization: This introduces a conservative error of 1– 10%; the Kirchhoff correction is recommended for ΔRH > 60%. For typical museum environments with ΔRH < 40%, the error is negligible (1–2%).
- Static loading: Fatigue and cyclic loading are not considered. The model provides a conservative upper bound for cyclic loading: if the static analysis indicates “safe,” the object is safe; if it indicates “unsafe,” fatigue may accelerate cracking. Extension to cyclic loading would require fatigue-modified and cumulative damage models.
- Bimaterial calibration: The calibration is valid for β ∈ [0.28,0.68]; the studied species have effective βeff = 0.103–0.251. The MAPE = 3.8% suggests that the extrapolation is robust, but recalibration is recommended for β < 0.1.
- Delamination: This is not addressed and requires a cohesive-zone formulation. Delamination would reduce stress transfer and increase tc (conservative) but would introduce new failure modes.
- Path dependence: The stress evolution in wood–gesso bilayers is path-dependent because of sorption hysteresis, nonlinear hygroscopic expansion, and the moisture-dependent elastic modulus. For arbitrary RH histories, the framework would need to be reformulated using convolution integrals, history-dependent GAB parameters, and memory variables.
- Aging model: This model is speculative, requires experimental calibration, and should not be used for conservation decisions.
2.11. Summary of Investigation Methods
Table 3 summarizes the analytical methods used at each stage of the framework, together with the manuscript section in which each is developed.
Table 3: Summary of investigation methods used in this study.
| Method | Description | Section |
|---|---|---|
| GAB sorption isotherm | Moisture content–RH relationship | Section 2.2.2 |
| Fourier eigenfunction expansion | Analytical solution of transient diffusion | Section 2.3.2 |
| Kirchhoff-mean diffusivity | Error correction for large RH steps | Section 2.3.3 |
| Modified Timoshenko bilayer formula | Calculation of the hygroscopic mismatch stress | Section 2.4.2 |
| Hutchinson–Suo perturbation expansion | Bimaterial correction factor Φ | Section 2.5.1 |
| ASED criterion | Fracture-mechanics assessment | Section 2.5.4 |
| Separation Line method | Safe/unsafe crack-depth threshold | Section 2.5.6 |
| Sensitivity analysis | Parameter perturbation study | Section 2.7.2 |
| FEM benchmark comparison | Validation (MAPE = 3.8%) | Section 2.5.7 |
| Literature validation | Comparison with independent experimental data | Section 2.8.1 |
3. Results and Discussion
3.1. Sorption Isotherms and Moisture Capacity
Figure 2 plots the GAB adsorption isotherms for pine wood, poplar, and gesso at T = 21 °C. All three exhibit the Type II shape typical of wood, and for pine, the moisture content rises from approximately 5% at RH = 10% to 25% at RH = 90%. The most notable feature is the difference between wood and gesso: at RH = 50%, pine holds approximately 12% moisture, whereas gesso takes up only 2.4% (see Table 1).
Figure 3 shows the specific moisture capacity CM(RH) and the effective diffusivity DRH for pine wood; both vary across the normal RH range. As shown, variation in DRH leads to the characteristic diffusion time τ1 ranging from 17 days at high RH to 673 days at low RH.
3.2. Diffusion Timescales and Interface RH Evolution
As shown for pine in Figure 4, for example, ranges from 17 days to 673 days.
Figure 5 and Figure 6 show the evolution of relative humidity with depth and at the interface. At short times (t < 10 days), the moisture front is confined to the outer surface region, while the gesso–wood interface (dashed line) remains at the initial RH. Between 10 and 50 days, the front reaches the interface, and at 100 days the whole wood reaches the new equilibrium.
Figure 6 presents the gesso–wood interface relative humidity versus time, while Figure 5 presents the relative humidity profiles with depth.
3.3. Hygroscopic Mismatch Stress
Figure 7 presents κ(RH). Two factors drive this order: the hygroscopic expansion coefficient αRH and the wood’s elastic modulus Ew.
Species-specific thresholds (Table 1) are as follows: pine (39.1%), lime (45.9%), spruce (59.8%), and poplar (62.9%). Objects stored below 20% RH require additional protection before exposure to 65% RH. For poplar, a step from 20% to 65% is safe (ΔRH = 45%<62.9%). Thus, the standard is conservative for poplar but not for pine, which justifies species-specific guidelines.
The time-evolving gesso stress (Figure 8) shows the sigmoidal growth pattern that is characteristic of diffusion-driven stress. For example, a step from 50% to 80% (ΔRH = 30%) generates a stress that plateaus at approximately 14 MPa, below the critical stress of 17.8 MPa; this scenario is permanently safe. A step from 30% to 90% (ΔRH = 60%) generates a stress that exceeds the critical value after approximately 40 days and continues rising to 30 MPa. The stress evolution is monotonic because the wood only swells; cyclic drying would introduce unloading and fatigue effects that are not considered here.
3.4. Averaged Strain Energy Density and Separation Line
Figure 9 plots the bimaterial correction factor Φ(η,α,β) for lime wood (β = 0.235) over thickness ratios α = 2 to 20, and several points are worth noting. Φ reaches a minimum at approximately η = 0.45; cracks located near the middle of the gesso layer experience the least amplification of the stress intensity factor, presumably because they are far enough from both the free top surface (which amplifies SIF) and the stiffer wood interface (which shields the crack tip) that neither effect dominates. As α increases, the wood layer becomes thicker.
Figure 9b compares the bimaterial correction factor for various woods, while Figure 10 shows the ASED versus the normalized crack depth η.
Using the Separation Lines in Figure 11 for a given panel α = L1/tg and species, one reads off η*(α), measures the actual crack depth hn, and computes η = hn/tg.
3.5. Critical Exposure Times
Figure 12 shows versus the different RH transitions. Once ΔRH ≥ 35% (RHf ≥ 40%), however, the ASED crosses at a time that depends on the species considered.
Each curve follows an S-shape, which mirrors the underlying diffusion: moisture penetrates slowly at first, so remains low; once the moisture front reaches the gesso–wood interface, rises rapidly; and then, as the wood approaches equilibrium at the new RH, levels off. tc itself is the point at which a given curve crosses the dashed line at . A rise from 5% to 50% (ΔRH = 45%), for instance, gives tc≈115 days, whereas a rise from 5% to 90% (ΔRH = 85%) takes only about 45 days.
Figure 13 plots tc versus ΔRH. For example, ΔRH = 40% from RH i = 5% gives tc ≈ 40 days, whereas from RH i = 50% gives tc ≈ 63 days.
Figure 14 plots tc for all 51 cases summarized in Table 4.
Table 4: Critical exposure times for selected RH transitions (pine, T = 21 °C, L1 = 40 mm, tg = 1 mm, η = 0.20).
| RHi [ %] | RHf [ %] | ΔRH [ %] | σmax [MPa] | tc [h] | tc [Days] | Status |
|---|---|---|---|---|---|---|
| 5 | 90 | 85 | 40.9 | 1090 | 45.4 | UNSAFE |
| 10 | 90 | 80 | 36.4 | 1225 | 51.0 | UNSAFE |
| 30 | 65 | 35 | 15.9 | >4320 | >180 | SAFE |
| 30 | 90 | 60 | 30.5 | 1191 | 49.6 | UNSAFE |
| 50 | 65 | 15 | 6.8 | >4320 | >180 | SAFE |
| 50 | 90 | 40 | 23.8 | 1521 | 63.4 | UNSAFE |
| 50 | 95 | 45 | 28.0 | 1149 | 47.9 | UNSAFE |
3.6. Species Comparison and Ranking
Figure 15 plots the critical exposure times for five initial RH values listed in Table 5.
Table 5: Analytically predicted unsafe crack-depth fractions Ψ [ %] for each species and boundary condition. Pine is omitted here because it is treated as the primary worked example in Table 4, which reports its full RH-dependent unsafe-fraction trajectory in detail.
| Species | BC1 Analytical | BC1 FEM | BC2 Analytical | BC2 FEM |
|---|---|---|---|---|
| Lime | 39–82% | 40–84% | 51–89% | 52–91% |
| Spruce | 40–49% | 41–50% | 51–62% | 52–63% |
| Poplar | 47–53% | 48–54% | 59–66% | 60–67% |
Lime exhibits the widest range of unsafe depths (39–82% uniaxial, 51–89% biaxial), reflecting its intermediate vulnerability. Poplar exhibits the narrowest range (47–53% uniaxial, 59–66% biaxial), confirming its safer status. The biaxial loading condition consistently yields higher unsafe fractions, emphasizing the importance of considering biaxial stresses in panel paintings.
3.7. Validation Against Literature
Figure 16 compares the analytical ASED with published FEM benchmarks. The MAPE of 3.8% (maximum error 5.3%) confirms that the semi-analytical ASED formulation accurately captures the crack-tip strain energy density across the full range of the parameters studied. The systematic conservative over-prediction (all analytical values being slightly higher than FEM) is desirable for conservation applications because it ensures that the framework does not underestimate the damage risk. The validation covers the lime, spruce, and poplar species, confirming that the bimaterial correction factor Φ calibrated for β ∈ [0.28,0.68] is robust for the species studied (βeff = 0.103–0.251).
The sensitivity analysis (Figure 18) demonstrates that the framework is robust to typical uncertainties in material parameters. Perturbations of ±10% in Ew and ±5% in GAB parameters Vm,c,k result in changes in tc of only 3–5% and 2–3%, respectively. No safe/unsafe reclassification occurs under any perturbation, confirming that the critical thresholds remain well separated from the uncertainty bands. This robustness is important for practical application because material properties vary within these ranges even among specimens of the same species.
The linearization error quantification (Figure 19) confirms that the midpoint linearization introduces a conservative under-prediction of tc of at most 10.2% for ΔRH ≤ 50%. The Kirchhoff-mean diffusivity αK (Kirchhoff-mean equation) reduces this error to below 3% in all cases studied. The conservative nature of the error (always under-predicting tc) is desirable because the model never misidentifies an unsafe scenario as safe.
3.8. Quantitative Sustainability Implications
The framework directly supports sustainable conservation decisions through several mechanisms:
Energy savings: The finding that 30% → 65% RH is safe for all species supports the use of winter setpoints as low as 30% RH. For a medium-sized museum (5,000 m2), reducing the winter RH setpoint from 50% to 30% could reduce heating energy consumption by an estimated 10–20%, corresponding to annual savings of approximately 50–100 MWh and CO2 emission reductions of 20–40 tonnes per year (based on typical HVAC energy-use data from [27]). This represents a significant contribution to sustainability goals without compromising the safety of the collection.
Material lifetime extension: By providing quantitative tc predictions, the framework enables targeted intervention before damage occurs, thereby extending the service life of panels and reducing the need for energy-intensive conservation treatments (which can require up to 5–10 times the embodied energy of preventive conservation). The early warning provided by tc allows museums to adjust environmental conditions before cracks initiate, thereby avoiding costly and resource-intensive restoration.
Species selection guidance: The ranking of the species by their vulnerability (Table 1: pine 39.1% ΔRH_crit, lime 45.9%, spruce 59.8%, poplar 62.9%) enables informed material selection for new commissions and replacements, favoring poplar in high-risk environments. This reduces the frequency of replacement and the associated environmental footprint of producing new panels.
Reduced replacement cycles: By enabling proactive conservation, the framework extends the functional lifetime of the panels, thereby reducing the demand for new materials and the energy associated with their production, transport, and installation.
3.9. Comparison with EN 16893:2018
The comparison between the present framework and EN 16893:2018 reveals several important insights. The standard’s 30–65% RH band is safe for objects stored above 20% RH, because a step from 30% to 65% (ΔRH = 35%) is safe for all species (pine: ΔRHcrit = 39.1%; poplar: 62.9%). However, for objects stored below 20% RH, the standard may prove insufficient. For example, a pine panel stored at 10% RH and exposed to 65% RH (ΔRH = 55%) would be unsafe (critical threshold 39.1%). This is particularly relevant for heritage objects in winter conditions, when RH may drop below 20%.
It is recommended that the standard be revised to: (1) specify a maximum ΔRH of 30% for pine panels, (2) specify a maximum of 40% for poplar, and (3) include a warning for objects stored below 20% RH before exposure to 65% RH. The finding that 30%→65% is safe for all species supports energy-efficient winter setpoints (30% RH), which could reduce heating energy consumption by 10–20% in museums and contribute to sustainability goals.
3.10. Practical Implementation and Significance
The Separation Line method (Figure 11) enables simulation-free safe/unsafe screening without the need for RH history modeling. Given the geometry of a panel (α) and its species, one can read the threshold crack-depth fraction η*. This is particularly useful for conservation surveys where crack patterns have been documented but RH data are not available. The critical exposure times tc provide the time available for intervention, enabling the prioritization of conservation actions. The species ranking guides material selection for new panels and the retrofitting of existing objects.
The framework supports energy-efficient environmental management by confirming that a winter setpoint of 30% RH is safe for a spring transition to 50–60% RH, provided that ΔRH < ΔRHcrit. This finding could reduce heating energy consumption in museums by 10–20% while maintaining collection safety, supporting sustainability objectives.
3.11. Limitations and Future Directions
The key limitations of the present framework include the following:
- One-dimensional diffusion: Lateral moisture flux near panel edges is neglected. For panels with lateral dimensions >5× the thickness, the error is <10% in the central region, whereas for small panels the model may over-predict tc by 10–20%.
- Midpoint linearization: This introduces a conservative error of 1–10%, which is acceptable but can be reduced using the Kirchhoff correction for ΔRH > 60%. For typical museum environments (ΔRH < 40%), the error is negligible.
- Static loading: Fatigue and cyclic loading are not considered. The model provides a conservative bound for cyclic loading: if the static analysis indicates “safe,” the object is safe; if it indicates “unsafe,” fatigue may accelerate cracking. Extension to cyclic loading requires fatigue-modified and cumulative damage models (Miner’s rule).
- Bimaterial calibration: The calibration is valid for β ∈ [0.28,0.68]; the studied species have effective βeff = 0.103–0.251. The MAPE = 3.8% suggests that the extrapolation is robust, but recalibration is recommended for β < 0.1.
- Delamination: This is not addressed and requires a cohesive-zone formulation. Delamination would reduce stress transfer and increase tc (conservative) but would introduce new failure modes.
- Path dependence: The stress evolution is path-dependent because of sorption hysteresis, nonlinear hygroscopic expansion, and the moisture-dependent elastic modulus. For arbitrary RH histories, the framework would need to be reformulated using convolution integrals, history-dependent GAB parameters, and memory variables.
- Aging model: This model is speculative, requires experimental calibration, and should not be used for conservation decisions.
- Experimental validation of tc: Direct validation of the predicted critical exposure times is needed using controlled RH chamber tests with real panel painting samples or well-characterized mock-ups.
Future work should include the following:
- Experimental validation of tc: Controlled RH chamber tests with real panel painting samples or well-characterized mock-ups should be performed to provide direct validation of the coupled time-dependent prediction.
- Extension to cyclic loading (fatigue): Fatigue-modified and cumulative damage models (e.g., Miner’s rule) should be validated against fatigue data from Refs. [19,20].
- 3D modeling of edge effects: For small panels (aspect ratio < 5), lateral diffusion should be incorporated to improve accuracy.
- Incorporation of delamination: Cohesive-zone formulations should be used to model interface failure.
- Arbitrary RH histories: The single-step diffusion solution should be replaced with a convolution integral using the Green’s function, incorporating sorption hysteresis through history-dependent GAB parameters and tracking the stress history for the path-dependent elastic modulus.
- Calibration of the aging model: The model should be calibrated with experimental data on paint-layer permeability and toughness evolution.
4. Conclusions
A fully analytical, validated three-stage framework for predicting the onset of hygroscopic damage in wooden panel paintings has been derived, tested against independent experimental data, and applied to four European wood species commonly used in heritage objects. The framework unifies transient moisture diffusion (Stage I-A), bilayer hygroscopic mismatch stress (Stage I-B), and fracture mechanics through the averaged strain energy density criterion (Stage II) into a closed-form, simulation-free tool that delivers critical exposure times (Stage III) for any given transition in relative humidity.
The main findings are as follows:
- Species-specific vulnerability: Critical ΔRH thresholds range from 39.1% (pine) to 62.9% (poplar), demonstrating that the widely used EN 16893 standard (30–65% RH) is safe for all species when starting from 30% RH but may be unsafe for pine and lime when starting from below 20% RH.
- Energy savings: The validation that 30% → 65% RH is safe for all species supports winter setpoints as low as 30% RH, potentially reducing heating energy consumption in museums by 10–20% (50–100 MWh/year for a medium-sized museum) without compromising the safety of the collection.
- Simulation-free screening: The Separation Line method η*(α) enables rapid safe/unsafe classification from crack-depth measurements alone, without requiring RH history data or numerical simulations.
- Robustness: The framework shows a MAPE = 3.8% against FEM benchmarks and is robust to typical uncertainties in material properties (±3–5% variation in tc for ±10% Ew perturbations).
- Sustainability: The species ranking and quantitative tc predictions enable informed material selection, targeted intervention, and extended service life, thereby reducing replacement cycles and energy-intensive conservation treatments.
The framework provides conservators, museum managers, and heritage scientists with a practical, validated tool for predicting humidity-induced damage, thereby enabling proactive conservation and sustainable environmental management.
List of Abbreviations
| Roman Symbols | |
| Symbol | Description |
| a | Radius of the penny-shaped crack (0.1 mm) |
| CM | Specific Moisture Capacity (∂MC/∂RH) |
| DRH | Effective Moisture Diffusivity |
| Ew,Eg | Elastic Moduli of Wood and Gesso |
| hn | Crack Depth from the Gesso–wood Interface |
| L1 | Wood Support Thickness (40 mm) |
| MC | Moisture Content (%) |
| R0 | Radius of the control volume for ASED (0.01 mm) |
| RH | Relative Humidity (%) |
| t | Time |
| tc | Critical Exposure Time (in days or hours) |
| tg | Gesso Layer Thickness (1 mm) |
| Vm,c,k | GAB Isotherm Parameters |
| x | Spatial Coordinate (through the thickness) |
| Greek Symbols | |
| Symbol | Description |
| α | Layer Thickness Ratio (tw/tg, range 2–20) |
| αϕ,RH | Tangential Hygroscopic Expansion Coefficient per unit RH |
| αm,αK | Midpoint and Kirchhoff-mean diffusivities |
| β | Elastic Mismatch Ratio (Ew/Eg, range 0.28–0.68) |
| βeff | Effective Elastic Mismatch Parameter based on plane-strain moduli |
| ΔRH | RH Step Change (RHf − RHi) |
| ϵc,ϵn | Compatibility Strain and Nominal Strain |
| η | Normalized Crack Depth (hn/tg, range 0.20–0.80) |
| κ | Mismatch Stress Coefficient (σg/ΔRH) |
| λn | Eigenvalue ((2n − 1)π/(2L1)) |
| νg | Poisson’s Ratio of Gesso (0.25) |
| ρ | Normalised Crack Radius (a/tg = 0.10) |
| σg | In-plane Gesso Stress |
| σcrit | Critical Gesso Stress (17.8 MPa) |
| τ1 | Characteristic Diffusion Timescale (17– 673 days) |
| Φ | Bimaterial Correction Factor |
| Φsf | Safety Factor (0.1, applied to critical ASED) |
| Averaged Strain Energy Density (ASED) | |
| Critical ASED Threshold (1.012 N/mm2) | |
| Acronyms | |
| Acronym | Description |
| ASED | Averaged Strain Energy Density |
| FEM | Finite Element Method |
| GAB | Guggenheim–Anderson–de Boer (sorption isotherm) |
| MAPE | Mean Absolute Percentage Error (3.8%) |
| RH | Relative Humidity |
| SIF | Stress Intensity Factor |
Availability of Data and Materials
The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.
Consent for Publication
Not applicable.
Conflicts of Interest
The author declares no conflicts of interest.
Funding
The study did not receive any external funding and was conducted using only institutional resources.
Acknowledgments
The author thanks the editorial team for the constructive feedback on this manuscript.
AI Declaration
Grammarly (Grammarly, Inc., San Francisco, CA, USA) was used for basic grammar and spelling checks. The author takes full responsibility for the accuracy, originality, integrity, and scientific content of the manuscript.
Appendix A. Derivation of the Semi-Analytical Correction Factor Φ
The correction factor Φ(η,α,β) is derived from a first-order Hutchinson–Suo boundary-perturbation expansion [6], and its functional form follows analytically:
The term C1(β) ⋅ η captures the free-surface amplification as the crack approaches x = tg, and the term C2(β) ⋅ (1 − η)2/(1 + α) captures the interface shielding as the crack approaches x = 0. The prefactors C1(β) = 0.083(β − 1)/(β + 1) and C2(β) = 0.146(β − 1)/(2β + 1) were calibrated against FEM data [7,21] and are valid for β ∈ [0.28,0.68]. Recalibration is required for species outside this range.
Appendix B. Linearization Error Bound
For ΔRH ≤ 40%, the errors remain below 8%. For ΔRH = 60%, the bound reaches ∼18%, which is consistent with the 10.2% maximum error observed in tc. The Kirchhoff-mean αK reduces this error to below 3% in all cases studied.
References
- [1] Time, B. Hygroscopic Moisture Transport in Wood. Ph.D. Thesis, NTNU, Trondheim, Norway, 1998.
- [2] Wadsø, L. Water Vapour Transport and Sorption in Wood. Ph.D. Thesis, Lund University, Lund, Sweden, 1993.
- [3] Bratasz, Ł.; Koźłowska, A.; Kozłowski, R. Analysis of Water Adsorption by Wood Using the GAB Equation. Eur. J. Wood Prod 2012, 70, 445–451. [CrossRef]
- [4] Timmermann, E.O. Multilayer Sorption Parameters: BET or GAB Values?. Colloids Surf. A 2003, 220, 235–260. [CrossRef]
- [5] Riparbelli, L.; Mazzanti, P.; Manfriani, C.; Uzielli, L.; Castelli, C.; Gualdani, G.; Ricciardi, L.; Santacesaria, A.; Rossi, S.; Fioravanti, M. Hygromechanical Behaviour of Wooden Panel Paintings: Classification of Their Deformation Tendencies Based on Numerical Modelling and Experimental Results. Herit. Sci 2023, 11, 1–19. [CrossRef]
- [6] Hutchinson, J.W.; Suo, Z. Mixed Mode Cracking in Layered Materials. Adv. Appl. Mech 1991, 29, 63–191. [CrossRef]
- [7] Beuth, J.L. Cracking of Thin Bonded Films in Residual Tension. Int. J. Solids Struct 1992, 29, 1657–1675. [CrossRef]
- [8] Bartolucci, B.; De Rosa, A.; Bertolin, C.; Berto, F.; Penta, F.; Siani, A.M. Mechanical Properties of the Most Common European Woods: A Literature Review. Frat. Ed. Integrità Strutt 2020, 14, 249–274. [CrossRef]
- [9] Güntekin, E.; Akar, S. Influence of Moisture Content on Elastic Constants of Scots Pine. Drewno 2019, 62. [CrossRef]
- [10] Fleischhauer, R.; Kaliske, M. Multi-Physical Modeling and Numerical Simulation of the Thermo-Hygro-Mechanical Treatment of Wood. Comput. Mech 2022, 70, 945–963. [CrossRef]
- [11] De Backer, L.; Laverge, J.; Janssens, A.; De Paepe, M. Diffusion Coefficient and Sorption Isotherm of Early Netherlandish Panel Paintings. Wood Sci. Technol 2018, 52, 149–166. [CrossRef]
- [12] Allegretti, O.; Raffaelli, F. Barrier Effect to Water Vapour of Early European Painting Materials on Wood Panels. Stud. Conserv 2008, 53, 187–197. [CrossRef]
- [13] Anderson, T.L. Fracture Mechanics: Fundamentals and Applications, 3rd ed.; CRC Press: Boca Raton, FL, USA, 2005.
- [14] Sneddon, I.N. Stress Distribution Near a Crack in an Elastic Solid. Proc. R. Soc. A 1946, 187, 229–260. [CrossRef]
- [15] Berto, F.; Lazzarin, P. Recent Developments in Brittle and Quasi-Brittle Failure Assessment by Means of Local Approaches. Mater. Sci. Eng. R 2014, 75, 1–48. [CrossRef]
- [16] Lazzarin, P.; Zambardi, R. A Finite-Volume-Energy Approach to Predict Fracture at Notches. Int. J. Fract 2001, 112, 275–298. [CrossRef]
- [17] Lazzarin, P.; Berto, F.; Zappalorto, M. Rapid Calculations of Notch Stress Intensity Factors. Int. J. Fatigue 2010, 32, 1559–1567. [CrossRef]
- [18] Jamalabadi, M.Y.A. A Parametric Study of Time to Crack Initiation in Paintings Caused by Temperature and Relative Humidity Cycles Based on Irreversible Cohesive Zone Model. SN Appl. Sci 2021, 3, 438.
- [19] Rachwał, B.; Bratasz, Ł.; Krzemień, L.; Łukomski, M.; Kozłowski, R. Fatigue Damage of the Gesso Layer in Panel Paintings. Strain 2012, 48, 474–481. [CrossRef]
- [20] Zhang, R.; Taylor, A.C.; Charalambides, M.N.; Balint, D.S.; Young, C.R.T.; Barbera, D.; Blades, N. Time for Crack Initiation in Panel Paintings Under Fatigue. J. Cult. Herit 2023, 61, 23–31. [CrossRef]
- [21] Bratasz, Ł.; Akoglu, K.G.; Kékicheff, P. Fracture Saturation in Paintings Makes Them Less Vulnerable to Environmental Variations in Museums. Herit. Sci 2020, 8, 11. [CrossRef]
- [22] Antropov, S.; Bratasz, Ł. Development of Craquelure Patterns in Paintings on Panels. Herit. Sci 2024, 12, 89. [CrossRef]
- [23] Jamalabadi, M.Y.A. Effects of Nonlinear Moisture Coefficients of Expansion on Rectangular Craquelure of Oil Panel Paintings. Res. Sq 2024. [CrossRef] [PubMed]
- [24] Khalid, S.; Azad, M.M.; Kim, H.S.; Yoon, Y.; Lee, H.; Choi, K.-S.; Yang, Y. A Review on Traditional and Artificial Intelligence-Based Preservation Techniques for Oil Painting Artworks. Gels 2024, 10, 517. [CrossRef] [PubMed]
- [25] Gebhardt, C.; Konopka, D.; Börner, A.; Mäder, M.; Kaliske, M. Hygro-Mechanical Study of the Katharinenaltar. J. Cult. Herit 2018, 29, 1–9. [CrossRef]
- [26] Sobón, M.; Bratasz, Ł. Risk of Fracture in Massive Wooden Heritage Objects. Eur. J. Wood Prod 2022, 80, 1201–1213. [CrossRef]
- [27] American Society of Heating, Refrigerating and Air-Conditioning Engineers. ASHRAE Handbook—HVAC Applications, Chapter 24: Museums, Galleries, Archives, and Libraries; ASHRAE: Peachtree Corners, GA, USA, 2019.
- [28] EN 16893; Conservation of Cultural Heritage—New Sites and Buildings Intended for the Storage and Use of Collections. CEN: Brussels, Belgium, 2018.
- [29] Jamalabadi, M.Y.A. Diagnosing Craquelure: A Mechanistic Case Study. J. Mar. Biol. Aquascape 2026, 3, 1–18.
- [30] Foti, P.; Califano, A.; Gao, C.; Sepe, R.; Bertolin, C.; Berto, F. Critical Exposure Time for Panel Paintings Due to Change in Environmental Conditions. Mech. Mater 2025, 202, 105234. [CrossRef]
- [31] Hilger, T.; Holl, K.; Hörmann, M.; Pallas, L.; Brandt, J.; Bellendorf, P. Unfavorable Relative Humidity as a Cause of Deterioration – Risk Assessment for the Humidification of a Medieval Polychromed Wooden Panel in Historic Context. Heritage 2025, 8, 526. [CrossRef]
- [32] Lorenzo, R.; Mazzanti, P.; Helfer, T.; Manfriani, C.; Uzielli, L.; Castelli, C.; Santacesaria, A.; Ricciardi, L.; Rossi, S.; Gril, J. Exploring the Inner Hygro-Mechanical Behaviour of Historical Panel Paintings: A Novel Approach Using Digital Twins. Herit. Sci 2024, 12, 27. [CrossRef]
- [33] Riparbelli, L.; Mazzanti, P.; Helfer, T.; Manfriani, C.; Uzielli, L.; Castelli, C.; Santacesaria, A.; Ricciardi, L.; Rossi, S.; Gril, J. Modelling of Hygro-Mechanical Behaviour of Wooden Panel Paintings: Model Calibration and Artworks Characterisation. Herit. Sci 2024, 12, 5. [CrossRef]
- [34] Califano, A.; Foti, P.; Baiesi, M.; Sepe, R.; Berto, F.; Bertolin, C. Machine Learning and Numerical Simulations for Predicting Critical Crack Conditions in Wooden Panels. Int. J. Struct. Integr 2025, 16, 914–935. [CrossRef]
- [35] Stöcklein, J.; Kaliske, M. Thermo-Hygro-Mechanically Coupled Modelling of Wood Including Two-Phase Moisture Diffusion for Transient Simulation of Wooden Structures at Mechanical and Climatic Loads. Heat Mass Transf 2023, 59, 67–79. [CrossRef]
- [36] Gagliardi, C.; Riparbelli, L.; Mazzanti, P.; Fioravanti, M. A Critical Review of Methods and Techniques Used for Monitoring Deformations in Wooden Panel Paintings. Forests 2025, 16, 546. [CrossRef]
- [37] Grbovic, A.; Sedmak, A.; Petrovski, B.; Sghayer, A.; Sedmak, S.; Razavi, J.; Berto, F. Numerical Simulation of Fatigue Crack Growth in AA6156 T6 Panels. Procedia Struct. Integr 2022, 39, 786–791. [CrossRef]
- [38] Romanowicz, M.; Grygorczuk, M. Experimental Investigation of the Mode I Fracture Energy of Wood Using DIC and Secant Compliance Techniques and Its Numerical Validation. Int. J. Solids Struct 2026, 325. [CrossRef]