The team, led by Raj Kumar Mishra of IIT Roorkee’s Fire Research Laboratory and Mahesh Kumar Tiwari of GD Goenka University, published their findings in Frontiers in Thermal Engineering on 13 August 2025. Their motivation is rooted in hard lessons from industrial history. During the initial stage of the 2010 fire at Dalian’s oil depot in China, flames damaged electricity supply cables and distribution rooms, interrupting the low-voltage power supply needed to electrically shut off emergency tank valves. Manual valve operation was slow and dangerous, and firefighting efforts suffered enormously. The episode illustrated a chilling truth: in a fire, key cables connected to safety equipment become the weak point of the entire emergency system.

To probe this vulnerability under controlled conditions, the researchers designed a bench-scale experimental campaign around a 1.1-kV, four-core, XLPE-insulated, armored, PVC-sheathed power cable with an aluminum conductor. The cable had an outer diameter of 26.2 millimeters and a current rating of 70 amperes. Specimens were cut to 1.2 meters in length, with 100 millimeters stripped from each end, leaving an experimental length of 1,000 millimeters. Two precisely drilled holes reached the insulation and the aluminum core, where high-accuracy K-type thermocouples were inserted to track temperatures deep inside the cable. Additional thermocouples monitored the outer surface and the flame itself, while a step-up transformer and variac supplied rated current through copper bus bars throughout the test.

The experimental protocol unfolded in three phases. First came the current phase, in which the cable carried its rated load for 20 to 25 minutes until its temperature reached a steady state, mimicking normal operation. Then the pre-burning phase began: fuel from a reservoir pan, maintained at a constant 50-millimeter fuel level, ignited and bathed the cable in external radiation. Finally, the burning phase arrived, when the cable itself became a source of ignition. The team defined failure by two criteria: electrical failure, marked by a short circuit when insulation damage caused a breakdown, and mechanical failure, when the cable surface began smoldering, igniting, or flaming. Both failure modes were considered in every test.

Experiments were conducted at two average heat flux levels, roughly 15 kilowatts per square meter (13 to 15 kW/m2) and 20 kilowatts per square meter (17 to 20 kW/m2), achieved by adjusting the distance between the burner and the cable. For each heat flux, five fire-retardant coating thicknesses were tested, ranging from 0 to 0.8 millimeters, yielding ten experiments in total. The fire-retardant paint, or FRP, was prepared by mixing 10 grams of flame-retardant agent with 10 milliliters of fire-retardant paint, with epoxy thinner added to improve workability. A single coat achieves 90 to 100 microns of thickness and dries within 20 to 24 hours. The cured coating was characterized on a physical property measurement system, revealing an average thermal conductivity of 1.633 watts per meter-kelvin, a specific heat capacity of 1,690.581 joules per kilogram-kelvin, and a density of 1,272.308 kilograms per cubic meter. Coating thickness was verified with a CALTECH gauge accurate to 1.5 percent or 2 percent.

The results for uncoated cables are sobering. Under an average heat flux of 15 kilowatts per square meter, the bare cable failed after 1,897 seconds, with the aluminum core reaching 282.60 degrees Celsius and the outer sheath 353.90 degrees Celsius at the moment of failure. Raising the heat flux to 20 kilowatts per square meter nearly halved the survival time: the cable failed after just 991 seconds, with core and sheath temperatures of 223.16 and 310.16 degrees Celsius respectively. The counterintuitive drop in failure temperature at higher heat flux reflects the thermal impact of intense radiation, which degrades the insulation faster than heat can diffuse inward, so the sheath breaks down before the core heats fully.

Coatings changed the picture dramatically. At 15 kilowatts per square meter, failure time increased from 1,897 seconds for the bare cable to 2,017 seconds at 0.2 millimeters of coating, 2,140 seconds at 0.4 millimeters, 2,241 seconds at 0.6 millimeters, and 2,542 seconds at 0.8 millimeters, a gain of nearly 11 minutes. At 20 kilowatts per square meter, failure times rose from 991 seconds bare to 1,204, 1,300, 1,386, and 1,522 seconds across the same thicknesses. The ignition time increased almost linearly with coating thickness, and the coating also altered failure chemistry: instead of melting and dripping off, the treated insulation hardened, adhering to the cable as solid particles or falling off in layers, making the burning cable less hazardous than bare wire. The fire-retardant paint suppresses the intense heat flux directly striking the cable and restricts flame spread along its surface.

Using the measured data, the team developed empirical correlations for cable failure time as a function of heat flux and coating thickness, fitted by least-squares regression. The correlations agreed with experimental results within 2 percent variation and provided a conservative estimate of time to failure. The researchers also introduced a temperature reduction ratio, or TRR, defined as the difference between the steady-state surface temperature of uncoated and coated cables divided by the bare cable temperature at failure. For the 15 kilowatts per square meter case, the maximum experimental TRR of 0.1822 occurred at 0.8 millimeters of coating, suggesting a critical thickness of 0.8 to 1.0 millimeters. For the 20 kilowatts per square meter case, the maximum experimental TRR of 0.1813 appeared at 0.2 to 0.4 millimeters, indicating that the optimal coating thickness depends on the fire’s thermal stress.

To extend the bench-scale findings, the team built a three-dimensional numerical model in COMSOL Multiphysics, replicating the cable’s complex cylindrical geometry with four aluminum cores, concentric insulation, inner sheath, galvanized steel armor, and outer PVC jacket. The mesh consisted of approximately 1,063,432 extremely fine tetrahedral elements, with local refinement in high-gradient regions. The model coupled the Heat Transfer in Solids and Electric Currents physics interfaces, applying Fourier’s law of heat conduction and Ohm’s law for current flow, with Dirichlet conditions fixing temperatures at inlets and Neumann conditions imposing heat flux at boundaries. A steady current of 210 amperes was supplied for 1,470 seconds to reach steady state, after which heat flux was applied to one-quarter of the cable’s outer sheath, matching the burner geometry in the experiments. The simulation tracked four stages: steady-state operation, failure by short circuit, complete burning, and convective cooling, which took nearly 3,498 seconds to return the cable to its initial temperature.

The numerical validation delivered strong agreement with the experiments. For the uncoated cable at 15 kilowatts per square meter, the simulation predicted a failure time of 1,898 seconds, just one second off the measured 1,897 seconds, with core and sheath temperatures of 333.37 and 361.83 degrees Celsius. At 20 kilowatts per square meter, the simulated failure time was 990 seconds against the measured 991 seconds, with core and sheath temperatures of 212.73 and 256.45 degrees Celsius. Across all ten cases, the absolute mean deviation was 9.94 percent for the outer sheath and 14.19 percent for the core. The team notes that the correlations were fitted to a narrow range of heat flux values due to limited experimental data, and validating them across wider heat flux ranges is future work. Still, the benchmark data now available can inform material specification decisions for cable installations in high-risk areas, guide the development of more effective fire-resistant coatings, and ultimately help engineers ensure that the cables meant to keep a plant safe do not become the spark that endangers it.

Subject of Research: Fire Safety Engineering

Article Title: A comparative study of cable fire dynamics: bench-scale experiments and numerical simulations with and without fire-retardant coating

Article References: Mishra, R. K., Tiwari, M. K., Dasgotra, A., Sharma, A., Gupta, A., Meena, M. K., Kumar, R., & Sharma, P. K. (2025). A comparative study of cable fire dynamics: bench-scale experiments and numerical simulations with and without fire-retardant coating. Frontiers in Thermal Engineering, 5, Article 1560746. https://doi.org/10.3389/fther.2025.1560746

Image Credits: AI Generated

DOI: 10.3389/fther.2025.1560746

Keywords: power cable, fire-retardant coating, heat flux, nuclear power plant, COMSOL simulation, cable failure, comparative, cable, fire, dynamics, bench-scale, experiments

The challenge is formidable. Acetes chinensis individuals measure just three to four centimeters, and in the echograms produced by a scientific echosounder they may occupy only a handful of pixels. The team, led by Yanrui Qu and Sheng Chen of the Third Institute of Oceanography in Xiamen, found that the median shrimp bounding box in their dataset covered just 0.189 percent of the image area. Layer on high-density aggregations, overlapping echoes, reverberation, and background acoustic clutter, and conventional object-detection models struggle badly, missing weak targets or mistaking noise for shrimp.

Acoustic sensing is the right tool for the job because it does not depend on light. Coastal waters near industrial intakes are often turbid, rendering optical cameras nearly useless, while a broadband split-beam echosounder such as the SIMRAD EK80 can operate regardless of visibility. But traditional hydroacoustic analysis, which relies on echo integration and target-strength measurements, estimates aggregate biomass rather than identifying and counting individual animals. Deep-learning detection fills that gap, providing explicit spatial localization that can be combined with tracking for automated enumeration.

The researchers started with RT-DETR, a real-time detection transformer that predicts objects end-to-end without the non-maximum suppression step used by most CNN-based detectors. Its self-attention mechanism models long-range dependencies between target and background regions, which matters when shrimp echoes are embedded in heterogeneous acoustic noise. The baseline RT-DETR-L, initialized with COCO-pretrained weights, already achieved a mean average precision at an intersection-over-union threshold of 0.5, or mAP@50, of 0.818 on their echogram dataset.

Three targeted modifications pushed performance higher. First, the team added a high-resolution P2 detection layer. The original architecture predicts from feature levels P3 through P5, whose spatial resolutions drop as the network deepens, losing fine detail about tiny targets. The new P2 branch fuses upsampled P3 features with shallow backbone features, producing a 160-by-160 feature map for 640-by-640 input, four times the spatial locations of P3, so weak echo boundaries survive the downsampling process.

Second, the researchers integrated Efficient Multi-Scale Attention, or EMA, modules at three backbone stages and three neck stages. These modules adaptively reweight features along both channel and spatial dimensions, amplifying shrimp-related responses while suppressing strong but irrelevant background echoes such as scattering and sidelobe interference. Notably, ablation experiments showed that placement matters: inserting EMA before the Adaptive Intra-scale Feature Interaction module yielded the best results, with mAP@50 rising to 84.2 percent, compared with 82.9 percent when placed between the encoder’s two modules.

Third, the loss function was rebuilt. Because background pixels vastly outnumber shrimp pixels, the team adopted Focal Loss for classification, which down-weights easily classified samples so training focuses on difficult cases. For bounding-box regression they used Wise-IoU version 3, which applies a dynamic non-monotonic focusing mechanism that reduces the influence of low-quality samples on gradient updates. This matters for small boxes, where even a slight displacement of a predicted boundary causes a large drop in overlap with the ground truth. The two losses proved complementary: together they lifted mAP@50-95, a stricter localization metric, to 0.437.

The fully modified model reached a mean mAP@50 of 0.844 and mAP@50-95 of 0.437, with precision of 0.834 and recall of 0.841, all statistically significant improvements over the baseline in paired t-tests. Recall improved by nearly six percentage points, indicating the architecture is particularly good at recovering weak targets that would otherwise slip past. The gains came cheap: parameters rose only from 31.99 million to 32.21 million, and the model outperformed small-object specialists QueryDet and TridentNet while remaining competitive with heavyweight detectors like DINO and RT-DETRv2, which it beat on recall despite their higher mAP scores.

Detection alone, however, cannot count shrimp, because the same individual appears in many consecutive frames. The team therefore coupled their detector with DeepSORT, a tracking algorithm that combines a Kalman filter motion model with appearance embeddings from a Wide Residual Network, matching detections to trajectories via the Hungarian algorithm. Each confirmed trajectory counts as one shrimp. Across ten ten-minute video clips, DeepSORT achieved a counting accuracy of 83.32 percent, outperforming SORT and ByteTrack, which reached 74.88 and 81.08 percent respectively. Qualitative examples showed the improved model dropping a false-positive background response from a confidence of 0.42 to 0.18, below the detection threshold, while boosting a weak shrimp echo from 0.11 to 0.35, rescuing a missed detection.

The authors are candid about limitations. Their data came from a controlled seawater tank, and generalization to different species, depths, densities, and acoustic environments remains untested. Counting errors still arise from trajectory fragmentation, identity switches, and temporary target disappearance, and the added P2 processing slows inference relative to the baseline. Future work will incorporate temporal and multi-frequency acoustic information and more robust trajectory association. Still, the framework demonstrates that deep-learning detection can extend from fish schools to individual planktonic crustaceans, giving nuclear plant operators a practical, non-invasive way to monitor biological loading before a shrimp swarm becomes a shutdown event.

Subject of Research: Deep learning-based detection and counting of Acetes chinensis shrimp in EK80 acoustic echogram images for nuclear power plant intake monitoring

Article Title: Improved RT-DETR-based detection and counting of shrimp in EK80 acoustic echogram images

Article References: Qu, Y., Chen, S., Xue, R., & Zheng, S. (2026). Improved RT-DETR-based detection and counting of shrimp in EK80 acoustic echogram images. Frontiers in Marine Science, 13, Article 1917117. https://doi.org/10.3389/fmars.2026.1917117

Image Credits: AI Generated

DOI: 10.3389/fmars.2026.1917117

Keywords: RT-DETR, deep learning, acoustic echogram, shrimp detection, Acetes chinensis, EK80 echosounder, object detection, multi-object tracking, DeepSORT, nuclear power plant, marine monitoring, small object detection

The research focuses on so-called coherency functions, statistical tools that quantify how strongly ground motions at two separated points on the ground surface are correlated as a function of frequency and distance. When waves propagate through the heterogeneous Earth, scattering, differences in arrival times known as wave passage effects, and local site conditions all erode this correlation, particularly at high frequencies. Because engineers cannot know the exact ground motion at every foundation point in advance, they rely on empirical coherency models fitted to dense instrument arrays, such as the SMART-1 array in Taiwan and the Lotung large-scale seismic experiment, to generate statistically consistent sets of spatially varying ground motion time histories for analysis.

A central motivation of the study is to validate the spectral correction factors recommended in ASCE Standard 4-98, the American Society of Civil Engineers standard for the seismic analysis of safety-related nuclear structures, and the coherency model developed by Norman Abrahamson in 2005 for the Electric Power Research Institute. These corrections account for the fact that the motion actually experienced by a large foundation is smoother, and generally weaker at high frequencies, than the free-field motion recorded in the soil away from the structure. This averaging effect, sometimes called the foundation input motion reduction, depends on the size and stiffness of the foundation, the properties of the underlying soil, and the assumed degree of incoherency in the incoming wave field.

To carry out the investigation, the team implemented several widely used coherency functions in code_aster, the open-source finite element software developed by EDF for advanced structural analysis. The models compared include classical formulations such as the Harichandran-Vanmarcke model, the Hao-Oliveira-Penzien model, the Der Kiureghian model and the Abrahamson and co-workers empirical functions derived from the Lotung experiment. Using these functions, the researchers generated ensembles of spatially varying ground motions compatible with target free-field response spectra for both rock and soil site profiles, following established simulation techniques for multivariate stochastic processes.

In the first part of the study, the authors analysed a rigid square foundation subjected to spatially varying ground motion that includes wave incoherency. From the computed foundation motions they derived foundation response spectra, which describe the maximum response a single-degree-of-freedom oscillator would experience at the base of the structure. By comparing these spectra with the corresponding free-field response spectra, they determined reduction factors, that is, ratios quantifying how much the foundation motion is attenuated relative to the free field at each frequency. These numerically derived factors were then benchmarked against the spectral corrections prescribed in the EPRI 2005 methodology and ASCE Standard 4-98.

The results reveal a striking sensitivity: the foundation response spectra can change considerably depending on which coherency model and which ground motion parameters are assumed. Because each empirical model was calibrated on different arrays, site conditions and frequency ranges, they do not produce identical descriptions of the wave field, and these differences propagate directly into the computed reduction factors. For practitioners, this means that the choice of incoherency model is not a technical detail but a decision that can materially influence the seismic demand calculated for safety-critical structures, particularly in the frequency range above a few hertz where incoherency effects are strongest.

The second part of the paper extends the analysis to a full reactor building of a nuclear power plant, coupling the spatially varying excitation with a soil-structure interaction model. Soil-structure interaction refers to the mutual influence between the flexibility of the supporting soil and the dynamic response of the structure: soft soils can lengthen the effective period of the building system and dissipate energy through radiation damping, while the massive foundation filters and averages the incoming wave field. When the ground motion is also spatially incoherent, these two mechanisms interact, and the resulting structural response can differ substantially from what a conventional analysis with identical motion at all support points would predict.

The findings show that the effect of incoherency is remarkable for secondary systems within the reactor building that are sensitive to frequencies larger than 10 hertz. Secondary systems include piping, cable trays, equipment and other components anchored to the structure, whose response is governed by the floor motion spectra at their attachment points rather than by the ground motion itself. Because incoherency predominantly suppresses high-frequency content in the foundation input motion, floor response spectra at these higher frequencies can be reduced, which may ease the design burden for high-frequency equipment, but the magnitude of this reduction depends again on the coherency model selected.

The work was carried out within an international collaborative framework, with support from the German Academic Exchange Service (DAAD), the European Commission’s METIS program under Horizon 2020, and the German Federal Ministry for the Environment, Nature Conservation, Nuclear Safety and Consumer Protection through the CRUAS-19 project. This combination of academic and industrial partners reflects the practical stakes of the research: nuclear regulators and operators in Europe and elsewhere are increasingly requiring that wave incoherency be explicitly accounted for in site-specific seismic assessments, and reliable, validated modelling tools are essential for that task.

For the earthquake engineering community, the study delivers a clear message about uncertainty and standardisation. The spectral corrections embedded in ASCE Standard 4-98 and the EPRI methodology represent decades of accumulated empirical knowledge, but the new results show that they should be applied with an awareness of the sensitivity of the outcome to the underlying coherency assumptions. As dense seismic arrays continue to record data worldwide, and as site-specific coherency functions are developed for individual nuclear sites, the authors’ open-source implementation in code_aster offers a transparent pathway for engineers to test alternative models, quantify the spread of predicted foundation and floor response spectra, and ultimately make more robust safety decisions for the large semi-rigid structures that society depends upon most.

Subject of Research: Effect of spatial ground motion coherency models on soil-structure interaction analysis of large semi-rigid structures such as nuclear reactor buildings

Article Title: Influence of different spatial coherency models in soil-structure interaction analyses of large semi-rigid structures

Article References: Soyluk, K., Zouatine, M., Sadegh-Azar, H., Zentner, I., Kudawoo, D., & Khemakhem, A. (2026). Influence of different spatial coherency models in soil-structure interaction analyses of large semi-rigid structures. Bulletin of Earthquake Engineering. https://doi.org/10.1007/s10518-026-02664-w

Image Credits: AI Generated

DOI: 10.1007/s10518-026-02664-w

Keywords: spatial variability, soil-structure interaction, coherency model, nuclear power plant, reactor building, seismic analysis, ground motion, wave incoherency, foundation response spectra, ASCE Standard 4-98, code_aster, earthquake engineering