This study investigates the thermal behaviour of two opaque and one transparent thin intumescent coating at material level (coating only) and system level (coating applied at real scale on timber), and their effects on timber in fire. Micro-scale tests were conducted to examine the underlying mechanisms of intumescence and degradation for each coating individually, while bench-scale tests demonstrated how these behaviours translate to larger scales. Critical temperature and heat flux thresholds were identified at which the coatings begin to insulate the timber through the formation of the intumescent char layer, as well as those marking degradation of the layer and the reduction of its insulating efficacy. The findings highlight that coating type and thickness, heating conditions and exposure duration influence mass retention, swelling pattern, and the integrity of the intumescent char layer. Overall, the transparent coating exhibited lower durability than the opaque ones. It was also shown that, due to the similar temperature ranges of timber pyrolysis and coating swelling, timber degradation occurs close to the coated surface during the transient swelling process, resulting in a heated region of 15-20 mm with negligible mechanical properties by the end of swelling. Therefore, intumescent coatings provide insulation progressively rather than immediately. • Integrity of swelled coating is influenced by type, thickness, and heat exposure • Transparent coating swelled layer exhibited lower durability than opaque coatings • Coatings on timber substrate swell faster and at lower heat fluxes than on steel • Timber 15-20 mm from the coated surface has minimal strength by the end of swelling

A Co-designed deep learning framework embedding physical insights for precise prediction of critical heat flux in microchannel flow boiling

Understanding heat transfer under near-critical pressure conditions is essential for the safe design of thermal-hydraulic systems such as future supercritical water reactors. While these systems operate under supercritical conditions during normal operation, subcritical states may occur during start-up, shutdown, or loss-of-pressure accidents. Under such conditions, a boiling crisis may develop if the critical heat flux (CHF) is exceeded, resulting in a sudden deterioration of heat transfer and a sharp increase in heating surface temperature. This behavior poses a significant safety concern, as excessive wall temperatures can result in fuel cladding degradation or failure and compromise overall system integrity. Experimental investigations on the boiling crisis, the associated critical heat flux, and heat transfer under post-CHF conditions have been conducted for several decades, leading to the development of various predictive approaches. However, most studies have focused on pressure ranges relevant to conventional pressurized water reactors. In contrast, experimental data at reduced pressures in the range 0.7 < p r < 1 remain scarce. To address this gap, the present study provides a comprehensive dataset on CHF and post-CHF heat transfer obtained from an industrial-scale test facility operating in this pressure regime. The dataset comprises 176 fully documented experiments using water as working fluid. The systematic variation of experimental parameters enables an interpretation of the contributions of the physical mechanisms governing the onset of the boiling crisis and post-CHF heat transfer with respect to pressure, mass flux, inlet temperature, and heat flux. This approach allows the identification of parameters that favorably or unfavorably influence CHF and characterize their impact on post-CHF heat transfer, thereby supporting the development and validation of improved safety-relevant prediction methods for next-generation nuclear reactor concepts. • Data gap in literature identified for CHF and post-CHF heat transfer. • Scarce data available for reduced pressures (pr > 0.7). • Water dataset generated at pr > 0.7 using an industrial-scale test-rig. • 176 CHF and post-CHF experiments with fully documented datasets. • Parametric trends identified at high subcritical pressures.

Boiling surface stability is essential for ensuring the safety and long-term performance of boiling-based thermal systems. While previous studies have addressed nanofluid boiling under repeated operation, this work focuses on finding the concentration threshold required for surface stabilization and explains how the sequence of nanofluid concentrations and post-stabilization dilution governs the evolution of SiO₂-deposited surfaces during successive pool boiling-cooling cycles. Nanoparticle concentration (0.01–2 wt. %), particle size (20–30 and 60–70 nm), and concentration sequencing effects are investigated. The results show that low concentrations (≤0.5 wt. %) of SiO2 nanofluid fail to stabilize surfaces within the accessible experimental range, whereas stable and repeatable behavior is achieved only at concentrations ≥1 wt. %. A surface stabilized with 2 wt. % SiO2 nanofluid achieves critical heat flux (CHF) values of about 1300 kW/m² and 1250 kW/m² for 20–30 and 60–70 nm nanoparticles, respectively, with wall superheats of about 52 °C. CHF is enhanced by 18 % and 13 % for 20–30 nm and 60–70 nm nanoparticles compared to deionized water, and this enhancement is preserved even after reducing the nanofluid concentration. On the 2 wt. %-stabilized surface, the boiling heat transfer coefficient (BHTC) rises with nanofluid concentration, and larger particles (60–70 nm) provide higher BHTC, with enhancements of approximately 8 %, 11 %, and 15 % at 2 %, 1 %, and 0.5 wt. %, respectively. Successive boiling-cooling cycles delay the onset of nucleate boiling (ONB) by about 75–85 % compared with the first cycle. The findings define a concentration-dependent stabilization threshold and propose a practical dilute-after-stabilization approach to achieve durable nanofluid boiling performance.

Bayesian-optimized, feature-augmented deep ensemble for physics-guided critical heat-flux prediction with uncertainty quantification

Experimental study on pool boiling heat transfer coefficient and critical heat flux on flat silicon surfaces under various heaving conditions

A Machine learning method based on extended Kalman filter to predict critical heat flux (CHF) Occurring in narrow channels between fuel plates

This study presents a novel phase-change cooling strategy that synergistically integrates acoustofluidic bubble dynamics with nanoarray-coated micropin fins to enhance thermal performance. A multidisciplinary framework combines experimental measurements-heat transfer coefficient (HTC), heat flux, velocity fields, and pressure drop-across SS, S30-120, and S-nanosheet architectures with machine-learning models, including LASSO, Random Forest, and Deep Neural Networks. Statistical correlation analyses (Spearman rank and Kendall tau) and interpretability techniques (SHAP, Partial Dependence Plots, Symbolic Metamodeling, Double Machine Learning, and TCAV) provide both predictive accuracy and physical insight. Nanosheet-coated surfaces deliver substantial performance gains, achieving a 71.2% increase in critical heat flux and a 160.9% improvement in HTC compared with smooth surfaces, validated by a highly accurate DNN model (R² = 0.99, MAE = 0.01). SHAP analysis identifies heat flux as the dominant factor governing bubble dynamics, while nanoarrays enhance nucleation and optimize pressure drop to improve efficiency. Symbolic Metamodeling and Double Machine Learning confirm heat flux as the primary driver of heat transfer, with secondary contributions from velocity, which influences bubble residence time, and nanoarray density. Feature-importance results further show that S-nanorod structures and pressure drop strongly regulate nucleation and flow behavior, whereas microreactor parameters have minimal influence.

Numerical study on the effects of simple spacer grid on critical heat flux

Experimental investigation on heat transfer performance and temperature uniformity of two-phase cooling in microchannels with different cross-section geometries

Deep generative modeling provides a powerful pathway to overcome data scarcity in energy-related applications where experimental data are often limited. By learning the underlying probability distribution of the training dataset, deep generative models, such as the diffusion model, can generate high-fidelity synthetic samples that statistically resemble the training data. Such synthetic data generation can significantly enrich the size and diversity of the available training data, and more importantly, improve the robustness of downstream machine learning models in predictive tasks. The objective of this paper is to investigate the effectiveness of diffusion models for overcoming data scarcity in nuclear energy applications. By leveraging a public dataset on critical heat flux which covers a wide range of commercial nuclear reactor operational conditions, we developed a diffusion model that can generate an arbitrary amount of synthetic samples. Since a vanilla diffusion model can only generate samples randomly, we also developed a conditional diffusion model capable of generating targeted critical heat flux data under user-specified thermal-hydraulic conditions. The performance of the diffusion model was evaluated based on its ability to capture empirical feature distributions and pair-wise correlations, as well as to maintain physical consistency. The results showed that both the diffusion model and conditional diffusion model can successfully generate realistic and physics-consistent critical heat flux data. Furthermore, uncertainty quantification results demonstrate that the conditional diffusion model is highly effective in augmenting critical heat flux data while maintaining acceptable levels of uncertainty. • A diffusion model was developed to generate synthetic critical heat flux data for nuclear energy. • A conditional diffusion model was developed for targeted generation at user-specified conditions. • The synthetic data was found to be physically consistent with existing mathematical models.

Experimental and mechanistic study on critical heat flux of R134a in tube under inclined conditions

Data-Driven Dimensional Analysis of Critical Heat Flux (CHF) in Subcooled Flow: Discovery of a New Key Dimensionless Number

Abstract This study experimentally investigates the thermal–hydraulic performance of ultra-confined two-phase jet-impingement cooling using a distributed inlet–outlet nozzle manifold, a configuration that has received limited attention in prior work. It addresses a critical knowledge gap by providing the first experimental characterization of a confined two-phase distributed inlet–outlet jet architecture operating at an extreme confinement height of 0.33 mm (h/d = 0.66), representative of practical chip-level cooling constraints. Simultaneous measurements of critical heat flux (CHF), pressure drop and local impingement cavity pressure enable direct evaluation of thermal–hydraulic tradeoffs under strong confinement. Experiments were conducted with the low-GWP refrigerant R1233zd(E) over flow rates from 0.15 to 1.25 LPM and jet-array densities corresponding to non-dimensional jet-to-jet spacing of s/d = 4.0–10.0. The results show that CHF and thermal–hydraulic efficiency are governed by jet-array density in ultra-confined regime, where liquid momentum, vapor evacuation, and confinement-induced flow resistance are tightly coupled. Maximum CHF values approaching 270 W/cm2 are achieved at ultra-low pumping powers below 0.4 W. Direct impingement cavity pressure measurements reveal a transition in the CHF-limiting mechanism from thermally dominated dryout at low flow rates to hydrodynamically constrained operation at higher flow rates. A scaling relationship linking CHF to jet Reynolds number and jet-array spacing is established, providing quantitative design guidance for high-performance ultra-confined two-phase jet-impingement cooling systems.

Hierarchical micro-nano surfaces with multi-functional synergy for boiling two-phase thermal management

Critical heat flux on boiling induced self-assembly surfaces

In this paper, a spatially non-uniform electrode is newly designed to be coupled with non-uniformly pillar-structured surfaces for enhancing the pool boiling performance under reduced gravity. Such a novel electrode consists of a three-dimensional (3D) mesh skeleton whose part of the mesh elements are bent downward. A 3D phase-change lattice Boltzmann model with an electric field model is employed to investigate the pool boiling performance and the associated boiling enhancement mechanism. It is found that the pool boiling performance on the pillar-structured surface with a uniform electrode is related to the mesh density of the electrode, as sparse meshes limit the action region of the electric field, whereas dense meshes may block the bubble departure. However, coupling the pillar-structured surface with either the planarly or the spatially non-uniform electrodes can significantly enhance the boiling performance, and the optimum enhancement is achieved through the synergistic effects of the spatially non-uniform electrode and non-uniform surface structures, with the critical heat flux and maximum heat transfer coefficient being increased by 48.3% and 19.6%, respectively, compared with those achieved on a uniformly pillar-structured surface with a uniform electrode. The numerical results reveal that the synergistic effects can not only simultaneously expand the electric field's action region and create wide bubble departure channels but also generate a strong non-equilibrium electric force and maximize the pinching-off effect. Therefore, more rounds of bubbles are detached earlier and independently, which prevents the bubble coalescence and provides more replenishment paths of fresh liquid for boiling enhancement.

Battery pack jet fire modeling and vehicle fire spread simulation in Ro-Ro ships

Critical heat flux prediction: A new approach adapting multi physics-aided machine learning

Critical heat flux (CHF) is a limiting thermal phenomenon in high-pressure boiling systems, and mechanistic models are employed to predict CHF by capturing the underlying physical triggers of the boiling crisis. In this study, two mechanistic CHF models, the bubble crowding model (BCM) and the vapor sublayer dryout model (SDM) are evaluated against an extensive experimental CHF database at pressures from 10 to 20 MPa. A comprehensive literature review is first presented, covering historical and recent developments in mechanistic CHF modeling and validation. The physical bases and governing equations of the BCM and SDM are summarized, and a methodology is described for applying these models to predict CHF with appropriate input parameters and closure correlations. The experimental CHF data (264 data points in a uniformly heated 8-mm tube) are described in detail, including the test facility, procedures, and CHF detection criteria. Model predictions are then compared quantitatively. Overall, the BCM achieves a mean absolute prediction error of ~14.5% and captures ~74.6% of data within ± 20% error, whereas the SDM shows ~24.8% mean error with ~69% within ± 20%. After excluding two extreme outliers at near critical conditions, the SDM performance improves to ~17.6% mean error (70% within ± 20%). The BCM outperforms the SDM on average, especially at low vapor qualities [departure from nucleate boiling (DNB) regime], while both models degrade in accuracy as conditions approach annular flow (quality > 0.3). In the highest quality bin (up to x ~ 0.43), the SDM markedly overpredicts CHF for certain outlying cases, underscoring the limits of its mechanistic assumptions outside its intended regime. While BCM appears more robust under high-pressure DNB conditions, neither model alone can universally predict CHF across all flow regimes. For both models, the accuracy deteriorates significantly as the flow nears annular conditions, highlighting the need for regime-specific mechanistic modeling (or hybrid approaches) to cover all regimes.