The research, led by Partheeb Mostafiz, Debolina Debnath and Kazi Md Tanvir Anzum, focuses on fast-moving consumer goods with extremely short shelf lives, using packaged bread as its case study. Bread has a commercial shelf life of just 120 hours, and retail scanner evidence indicates that bakery lines exhibit the highest expiration rates of any perishable category, with more than 8% of perishable units expiring before sale and bakery products frequently exceeding 12%. Distributors of such products face a chronic dilemma: overstocking secures service levels but converts edible product into end-of-life waste, while understocking protects margin at the cost of lost sales and customer defection. The problem is sharpest in the traditional-trade channel that dominates South Asian retail, where neighbourhood outlets typically operate without electronic point-of-sale systems, real-time inventory visibility or formal demand forecasting, so replenishment quantities are set judgementally days before demand is realised.

The team’s answer is a dual-channel clearance mechanism they call the Near-Food Channel, or NFC. At a fixed audit point 72 hours after production, any inventory in excess of an optimised threshold, termed the Protection Level, is physically transferred to a separate discount retailer. Because clearance stock is spatially and commercially insulated from the full-price shelf, it cannot cannibalise fresh sales, the very outcome that has discouraged distributors from adopting markdown pricing in the first place. Existing markdown models typically assume that discounted units remain on the primary shelf, where they erode full-price revenue; the new design removes that penalty by construction rather than by pricing cleverness.

Determining how much inventory to protect required a stochastic mixed-integer non-linear programme, solved in GAMS using sample average approximation over 300 Monte Carlo scenarios, and then validated in a Rockwell Arena 14.0 discrete-event simulation of the full 120-hour product lifecycle. Crucially, the two behavioural parameters that govern the policy were estimated from primary data rather than assumed. A structured survey of 217 traditional-trade bread purchasers in Dhaka yielded a freshness decay penalty of 0.72, fitted by non-linear least squares to a power-decay model of willingness to pay, and a clearance-channel price elasticity of 2.85, obtained by log-log regression of conversion rates on relative price. Both parameters entered the models as normal distributions with a standard deviation of 0.05, so that estimation uncertainty propagated through every scenario and replication rather than being suppressed.

The empirical setting was a high-volume retail node in Jatrabari, Dhaka, comprising 20 neighbourhood outlets served by a single distributor. Daily footfall across the catchment was recorded, tested for normality and represented as a normal distribution with a mean of 1,040 customers and a standard deviation of 30, while individual inter-arrival times within the trading day were generated as exponential variates conditional on the realised daily total. The distributor buys each unit at 40 Bangladeshi taka and sells fresh product at 50 taka, the clearance retailer operates on a 10% margin, and reverse logistics cost 0.5 taka per transferred unit. Under the conventional baseline, the manufacturer grants a 2% wastage allowance that reduces the distributor’s effective unit cost to 39.2 taka, an arrangement the authors describe as an implicit subsidy for waste.

The numerical results are striking. In the conventional single-channel system, the model identifies a sharp waste threshold at 2,126 units, the mean quantity the fresh channel absorbs over the lifecycle: below it the entire order sells, and above it every additional unit ordered is wasted. At the maximum tested stocking level of 3,000 units, the baseline produces 874 wasted loaves, a 29.13% waste rate, and a distributor loss of 11,300 taka, with profit turning negative from 2,750 units onward. Both dual-channel scenarios, one applying a constant 30% markdown throughout the clearance window and the other a stepped 20% on Day 4 followed by a further 10% on Day 5, eliminated physical waste entirely across the full range of stocking levels from 1,700 to 3,000 units.

The financial consequences are equally significant. Under the constant markdown, distributor profit stabilises between roughly 19,400 and 20,900 taka at stocking levels where the conventional structure would have lost money, while the stepped schedule performs better still, delivering 20,346 taka at maximum overstock and generating 17% more gross clearance revenue, 23,210 taka against 19,810 taka. The advantage of the stepped policy stems from simple arithmetic: most clearance volume moves on Day 4 at only a 20% reduction, returning 36.36 taka per unit to the distributor rather than the 31.82 taka recovered at a 30% cut, and the deeper discount applies only to the residual. Total ecosystem value rises from 106,300 taka under the conventional structure to more than 142,000 taka under the stepped policy, with the gains drawn from waste that previously accrued to no one rather than redistributed between parties.

A one-at-a-time sensitivity analysis confirmed that the profitability benefit is robust. Doubling and even quadrupling the reverse-logistics cost, and raising the clearance retailer’s margin from 10% to 20%, left waste mitigation efficiency at 100% and reduced distributor profit by at most 7.7%, a consequence of the fact that the recovered salvage value exceeds the transfer cost by a factor of more than sixty at every markdown depth tested. The binding constraints proved to be demand-side rather than cost-side: a 10% reduction in catchment footfall was the single most damaging perturbation, cutting waste mitigation efficiency to 90%, while a 20% increase in the freshness decay penalty reduced it to 92.3%. Even under the worst perturbation, however, distributor profit remained positive, ranging from 5,483 to 23,973 taka against the conventional baseline loss of 11,300 taka at the same stocking level.

Beyond the numbers, the study carries a broader theoretical claim: waste reduction and profitability are complementary rather than competing objectives once the channel structure is right. Prior simulation-based studies achieved waste reductions of 15 to 30% through single-channel discounting and treated the residual as an irreducible cost of service. The present results suggest that this residual is largely an artefact of channel design, because physically separating clearance stock from full-price stock removes the cannibalisation penalty that bounds markdown policies. The framework also dissolves the distributor’s incentive to under-order: with the payoff function flattened and never turning negative, the ordering rule can shift from a conservative quantile of the demand distribution to a service-driven one, and the manufacturer recovers the full wholesale price of 40 taka per unit, worth approximately 2.92 million taka annually to a producer shipping 10,000 units per day.

The practical appeal of the mechanism lies in its low-tech implementability. Because the Protection Level can be expressed as a lookup table indexed by the initial order quantity, front-line warehouse staff can execute the audit and transfer from a printed schedule, requiring no real-time analytics, connectivity or point-of-sale integration, a decisive consideration in exactly the traditional-trade settings where data-intensive waste-reduction approaches have failed to take hold. The authors argue the framework transfers to packaged dairy, ready-to-eat meals, cut produce and pre-packed protein, subject to re-estimation of the behavioural parameters and adjustment of the audit timing. At policy level, they suggest that wastage allowances written into manufacturer-distributor contracts could be discouraged, and that an explicit regulatory category for near-expiry sale with mandatory date disclosure would reduce the legal uncertainty that currently deters clearance retailing. Given that retail accounts for roughly 13% of the 931 million tonnes of food wasted globally each year, a profitable, subsidy-free intervention of this kind offers emerging economies a low-capital complement to the cold-chain investments that have long dominated food-loss strategy.

Subject of Research: Stochastic dual-channel clearance routing to prevent food waste in perishable FMCG supply chains

Article Title: A stochastic dual-channel clearance strategy for food waste prevention in perishable FMCG supply chains

Article References: Mostafiz, P., Debnath, D., & Anzum, K. M. T. (2026). A stochastic dual-channel clearance strategy for food waste prevention in perishable FMCG supply chains. Cleaner Engineering and Technology, 34, Article 101310. https://doi.org/10.1016/j.clet.2026.101310

Image Credits: AI Generated

DOI: 10.1016/j.clet.2026.101310

Keywords: food waste, perishable supply chains, FMCG, markdown pricing, simulation-optimisation, discrete-event simulation, protection level, clearance channel, reverse logistics, cleaner production, Bangladesh, bakery

The technology addresses a persistent and underappreciated problem in livestock production. Rice husk, an abundant by-product of rice milling, is widely used as poultry bedding because it is cheap, lightweight, and absorbs moisture well. But during storage and use, bedding accumulates organic matter, fecal residues, and moisture that support microbial growth. Contaminated bedding can act as a long-term environmental reservoir for foodborne pathogens, facilitating transmission between production cycles and throughout the poultry supply chain. Conventional sanitation relies on chemical disinfectants such as quaternary ammonium compounds, aldehyde-based treatments, and oxidizing agents, yet these often penetrate bulk biomass poorly, work unevenly in porous substrates, and lose effectiveness under high organic loading, while raising concerns about residues, worker exposure, and environmental impact.

The new system, described in the journal Cleaner Engineering and Technology, is a trailer-mounted, mobile processing platform designed for continuous on-site operation. Its core components include an insulated sanitization chamber, a screw-assisted biomass transport mechanism that lifts and cascades rice husk particles through the treatment zone, an LPG-fueled gas burner generating hot air at 350 to 400 degrees Celsius, a hot-air mixing chamber, forced-air and recirculation blowers, and a cyclone dust separation unit. The screw-assisted cascading design is central to the system’s performance: by continuously lifting and redistributing particles, it repeatedly exposes fresh surfaces to the heated airflow, minimizing localized cold spots and thermal dead zones within the porous biomass bed.

A key engineering innovation is the airflow recirculation architecture. Rather than exhausting heated air after a single pass, the system captures hot air exiting the chamber, passes it through cyclone separators to remove suspended dust, and redirects it back into the heating circuit. This recirculation loop achieved an efficiency of approximately 86.7 percent, substantially improving thermal retention, stabilizing chamber temperature under transient loading, and reducing fuel consumption. Under steady-state conditions, the sanitization chamber held temperatures of roughly 115 to 125 degrees Celsius while rice husk exited at 95 to 100 degrees Celsius, with chamber temperature variation held to about plus or minus 2.1 degrees Celsius.

Residence time proved to be a critical operating variable. By tuning screw rotational frequency between 50 and 60 hertz, the researchers achieved residence times of 15 to 17 minutes, which they identified as the optimal window balancing microbial inactivation against throughput. At these settings the system processed 2.0 to 2.5 tonnes of rice husk per hour with thermal utilization efficiency peaking near 87.6 percent. Shorter residence times below 10 minutes risked ineffective sanitization, while longer exposures reduced productivity without meaningful gains. Statistical analysis confirmed that both temperature and residence time significantly influenced microbial reduction, with their interaction significant at p less than 0.01.

The microbiological validation was unusually rigorous for an industrial study. Throughout three years of routine monitoring, approximately 67 samples per month were collected from production lots of about 15 tonnes each, and formal annual validations compared 50 samples before treatment with 50 after. All analyses were performed at an ISO/IEC 17025-accredited laboratory, using selective enrichment, culture on Xylose Lysine Deoxycholate agar, and confirmatory identification by MALDI-TOF mass spectrometry. Under optimized conditions, contamination frequency fell from 16 percent to 4 percent, an estimated 75 percent reduction in positive detections. In controlled laboratory experiments, log reductions of total bacterial count reached 6.2 to 6.4 log CFU per gram at 120 to 125 degrees Celsius, corresponding to inactivation efficiencies above 99.999 percent, and Salmonella was not detected under any tested condition.

Perhaps the most instructive finding concerned moisture. Feedstock water content emerged as a dominant constraint on performance, contributing an estimated 18.6 percent of the relative influence on lethality. At moisture levels of 8 to 11 percent, the system achieved log reductions of about 6.3, but at 18 to 20 percent moisture the reduction fell to roughly 2.7 log, a level the authors classified as ineffective. The explanation is thermodynamic: water evaporation consumes latent heat, diverting thermal energy away from microbial destruction and buffering the biomass temperature. Field validation during rainy conditions confirmed the pattern, with positive detections rising to 9 percent at high moisture and 14 percent when high moisture was combined with shortened residence time. The researchers argue that industrial thermal sanitization must therefore be governed by integrated heat-and-moisture frameworks rather than temperature-only criteria, and they recommend real-time moisture sensing and adaptive residence-time control as future upgrades.

The economics are compelling. Total treatment cost was estimated at 0.0083 US dollars per kilogram of rice husk, compared with 0.014 to 0.021 dollars for conventional centralized chemical sterilization, an operating cost reduction of roughly 40 to 60 percent and a benefit-cost ratio of 4.6. LPG consumption averaged 8.3 kilograms per tonne of treated biomass, translating to a fuel cost of about 0.0052 dollars per kilogram. The savings arose from multiple vectors: complete elimination of disinfectant procurement, which alone accounted for 38.4 percent of total savings; removal of dedicated chemical storage warehouses; simplified single-stage logistics replacing multi-stage transport to centralized facilities; and reduced transportation emissions from on-site processing. Over three years, cumulative savings exceeded 0.30 million US dollars, averaging about 100,000 dollars annually.

From a sustainability standpoint, the system embodies cleaner production and circular economy principles. By using thermal energy as the sole sanitizing agent, it removes chemical residues from bedding that might otherwise accumulate in soils when spent litter is applied as fertilizer, and it eliminates occupational exposure to formaldehyde and related compounds. The life-cycle analysis, conducted through a life-cycle thinking approach rather than a formal ISO 14040/14044 assessment, attributed 34.8 percent of sustainability benefits to chemical elimination, 26.5 percent to biomass reuse and valorization, and 11.6 percent to avoided transportation. The technology effectively upcycles a low-value agricultural residue into a standardized sanitary product, closing a resource loop that links rice milling, poultry farming, and eventual nutrient recycling through spent bedding.

The authors are careful to note limitations. The microbiological outcomes reflect the tested operating conditions and should not be read as evidence of complete pathogen elimination under all field scenarios, and the sustainability percentages are relative indicators rather than quantified greenhouse gas reductions. Future work will pursue a full cradle-to-grave life cycle assessment, quantitative microbial inactivation modeling using D-value and z-value kinetics, computational fluid dynamics optimization of the reactor, and predictive process control. Still, the demonstrated combination of scale, reliability, and economics marks a significant step. A system that processes nearly 8,000 tonnes of abrasive, silica-rich biomass over three years with greater than 95 percent availability shows that continuous thermal sanitization has moved beyond the pilot stage, offering poultry producers a practical blueprint for biosecurity that is simultaneously cleaner, cheaper, and more circular than the chemical status quo.

Subject of Research: Industrial-scale continuous thermal sanitization of rice husk bedding for chemical-free poultry biosecurity

Article Title: Industrial-scale continuous thermal sanitization of rice husk bedding for cleaner poultry production: Resource valorization, biosecurity improvement, and sustainability assessment

Article References: Sangpradit, K., Thoetrattanakiat, S., Sookyoo, W., Intarasuk, A., & Samseemoung, G. (2026). Industrial-scale continuous thermal sanitization of rice husk bedding for cleaner poultry production: Resource valorization, biosecurity improvement, and sustainability assessment. Cleaner Engineering and Technology, 34, Article 101304. https://doi.org/10.1016/j.clet.2026.101304

Image Credits: AI Generated

DOI: 10.1016/j.clet.2026.101304

Keywords: rice husk, poultry bedding, thermal sanitization, biosecurity, cleaner production, circular economy, pathogen inactivation, agricultural waste valorization, hot-air heating, Salmonella, LPG energy efficiency, sustainable livestock production

The study, published in the Journal of Industrial Ecology, introduces a multi-objective, multi-level cleaner production evaluation framework that does something traditional scorecards have long failed to do: it treats carbon reduction as a first-class criterion rather than an afterthought. Grounded in the principles of industrial ecology, the framework combines a full life-cycle perspective with a dedicated carbon dimension, forcing evaluators to look beyond the smokestack and consider impacts from raw material extraction all the way through to the final steel product.

Cleaner production, a concept that has guided industrial environmental policy for decades, traditionally emphasizes reducing pollution and conserving resources within factory boundaries. But as climate change has moved to the center of global concern, that narrow focus has become inadequate. The Fudan team, led by Shoubing Wang together with Yuxuan Zhang, Qi Wei and Jingru Hu, argued that evaluation systems built before the carbon era simply cannot capture the full environmental profile of a modern steel enterprise, which must now answer for greenhouse gas emissions across its entire value chain.

To construct the framework, the researchers wove together three complementary methods. A systematic policy review of China’s cleaner production regulations for the steel sector revealed how policy priorities have evolved through three distinct stages, from an initialization phase, through rapid development, and into the current era of innovation and carbon focus. A life-cycle inventory analysis then defined the system boundary and screened which indicators genuinely matter across the steel production chain. Finally, expert consultation refined and validated the resulting indicator set, ensuring the framework reflected both scientific rigor and practical industrial realities.

Scoring within the framework relies on two well-established analytical techniques. An improved analytic hierarchy process, a structured method for deriving weights from pairwise comparisons, determines how much each indicator contributes to the overall score. Fuzzy comprehensive evaluation then converts indicator data into performance grades, an approach well suited to environmental assessment where many measures are inherently imprecise or qualitative. The combination allows the framework to aggregate dozens of indicators spanning resource consumption, energy use, pollution, product management and carbon reduction into a single interpretable score.

The real test came at Baosteel, one of China’s flagship steel producers and a subsidiary of the Baowu Group, the world’s largest steelmaker. Applying the framework to real operational data, the researchers found that Baosteel’s cleaner production score climbed from 62.68 in 2014, a rating classified as Domestic Poor, to 86.39 in 2021, achieving the Domestic Advanced category. That leap of nearly 24 points over seven years represents a striking transformation, but the story behind the numbers proved even more valuable than the headline result.

The evaluation did more than certify improvement; it pinpointed precisely where the company still fell short. Two critical weaknesses emerged: carbon reduction and product life-cycle management. Even as Baosteel advanced on conventional pollution and resource metrics, the dedicated carbon dimension exposed gaps that a traditional assessment would likely have missed or underweighted. This diagnostic capability is arguably the framework’s most important contribution, because it transforms evaluation from a passive grading exercise into an actionable roadmap, generating specific recommendations for where investment and management attention should flow next.

The timing of this work could hardly be more significant. China’s steel industry, the world’s largest producer, has been formally incorporated into the national carbon emissions trading market, and the sector faces intense pressure under the country’s carbon neutrality pledges. For an industry responsible for a substantial share of global carbon dioxide emissions, the difference between a scorecard that treats carbon as peripheral and one that places it at the core could shape investment decisions worth billions. A framework that quantifies carbon performance alongside energy efficiency and pollution control gives regulators, investors and company managers a common language for assessing progress.

The researchers are careful about the limits of their findings. The full case study rests on a single Chinese enterprise, and they emphasize that the results require cautious interpretation until the framework is validated across broader industrial and geographical contexts. The framework is presented as a proof of concept, a preliminary tool that could, with further contextualization and testing, help steel companies navigate sustainable production under tightening carbon constraints. Data limitations also reflect commercial realities: the raw Baosteel data was provided under license by Baosteel Engineering & Technology Group Co., Ltd. and cannot be made public due to confidentiality agreements, although the team notes supporting data are available upon reasonable request.

Nevertheless, the implications reach well beyond one company or one country. As carbon border adjustments, emissions trading schemes and net-zero commitments spread across the global economy, heavy industries everywhere will need credible, standardized ways to measure and compare environmental performance. A life-cycle based evaluation framework with a dedicated carbon dimension offers a template that could be adapted to cement, aluminum, chemicals and other hard-to-abate sectors. If cleaner production evaluation once measured how little smoke a factory emitted, the next generation of scorecards will measure how deeply a company has rewired its entire metabolism for a carbon-constrained world, and China’s steel industry, the study suggests, has shown just how far that rewiring can go in a single decade.

Subject of Research: A life cycle based cleaner production evaluation framework with an integrated carbon dimension, applied to China's steel industry

Article Title: A life cycle based cleaner production evaluation framework integrating a dedicated carbon dimension: a case study of China’s steel industry

Article References: A life cycle based cleaner production evaluation framework integrating a dedicated carbon dimension: a case study of China’s steel industry. (n.d.). https://doi.org/10.1007/s44498-026-00155-4

Image Credits: AI Generated

DOI: 10.1007/s44498-026-00155-4

Keywords: steel industry, cleaner production, carbon reduction, life cycle assessment, industrial ecology, China, Baosteel, evaluation framework, carbon neutrality, multi-criteria decision analysis, fuzzy comprehensive evaluation, sustainable production

The research, led by Junjie Cai, Xijian Li and Shoukun Chen of Guizhou University, focuses on the 21,605 working face of the Qinglong Coal Mine in Guizhou, China. The team built a detailed multi-physics model of the coal seam and its gas drainage system using COMSOL software, simulating 200 meters of seam length, 150 meters of width and 3 meters of height with a single horizontal extraction borehole. Their model coupled gas flow, heat conduction, coal deformation and pressure distribution, drawing on Langmuir adsorption equations, cubic-law permeability relationships and thermo-poroelastic strain terms to describe how gas moves through coal as pressure, temperature and stress evolve.

From this physical foundation, the researchers designed twelve orthogonal simulation schemes varying four key parameters over 120 days of extraction: negative extraction pressure, borehole radius, temperature and coal seam permeability. The simulations revealed elliptical zones of depressurized gas around the borehole, with cumulative extraction volumes rising rapidly during the first 40 days before stabilizing. Critically, scenarios with enhanced permeability and altered temperature produced markedly higher extraction volumes, foreshadowing the sensitivity analysis that would later confirm permeability and temperature as the dominant controlling factors.

Armed with simulation output, the team turned to machine learning. They trained two models, the gradient-boosting algorithm XGBoost and the deep learning Long Short-Term Memory network, on the simulation data, using negative pressure, radius, temperature and permeability as inputs and gas extraction volume as the output. To squeeze out maximum performance, they optimized both models with four bio-inspired optimization algorithms: the Sparrow Search Algorithm, Particle Swarm Optimization, Grey Wolf Optimizer and the White Shark Optimizer, each of which tunes hyperparameters by mimicking the search behavior of animal collectives.

The results were striking. Individually, the models performed well; XGBoost, in particular, proved more robust than LSTM when training samples were limited, a known weakness of deep networks on small datasets. But the decisive breakthrough came from a hybrid architecture: an LSTM feature encoder feeding enhanced representations into XGBoost for final regression. Evaluated with the Entropy Weight Method across MAE, MAPE, MSE, RMSE and the coefficient of determination, the optimized XGBoost-LSTM hybrid achieved the highest score of any configuration. Relative to the base XGBoost model, its MAE, MAPE, MSE and RMSE fell by 72.64, 46.89, 95.77 and 79.44 percent respectively; against the base LSTM, the reductions reached 79.53, 33.39, 98.65 and 88.37 percent. The R² value climbed to 0.9998, effectively near-perfect agreement with observed extraction volumes.

When benchmarked directly against numerical simulation, the machine learning approach showed a clear advantage. Simulation error ranged between minus 7 and 7, while the best model, GWO-XGBoost, narrowed errors to between minus 3 and 1. The authors note that below a threshold extraction volume of 50 by ten-to-the-fourth cubic meters, predictions were tight and reliable; at higher volumes, sparse training samples caused error dispersion, identifying data scarcity as the chief limitation on generalization.

Explainability analysis using SHAP values reinforced the physical story. Permeability-enhancing schemes carried predictive importance two to two-and-a-half times that of other scenarios, confirming that increasing coal seam permeability, for example through hydraulic fracturing, liquid nitrogen cryogenic treatment, or shock-wave fracturing techniques, is the most effective lever for boosting gas yield. Temperature contributed indirectly by influencing permeability, while extraction pressure and borehole radius offered only modest gains under the tested conditions.

The implications extend well beyond prediction. Accurate gas volume forecasts allow mines to capture methane as fuel for power generation, heating and chemical synthesis rather than venting it, cutting greenhouse gas emissions while improving mine safety against coal and gas outbursts. The framework also supports coupled technologies in which methane substitution for fossil fuels and CO2 displacement with geological storage jointly deliver emission control, resource efficiency and carbon sequestration, precisely the demands of low-carbon coal production.

The authors are candid about limitations: the dataset derived from twelve orthogonal schemes and 120 days of monitoring, excluded factors such as moisture content, pore structure and in-situ stress, and optimization algorithms did not always improve every model. Future work will expand field data, enrich input parameters, develop interval and time-series prediction, and validate the system across different mines and geological settings, paving the way for genuinely intelligent, unmanned gas extraction in the service of a greener coal industry.

The dual character of coal seam methane has long shaped how the mining industry approaches drainage. On one hand, methane concentrations in underground workings must be kept below explosive thresholds, and sudden releases of gas and coal, known as outbursts, remain among the deadliest hazards in deep mining. On the other hand, the same molecule carries a calorific value comparable to natural gas and can displace coal in power generation with substantially lower carbon dioxide emissions per unit of energy. Capturing gas before, during, and after mining therefore converts a liability into an asset, which is why prediction accuracy carries such practical weight in drainage system design, borehole scheduling, and surface utilization planning.

The physical model underlying the study rests on well-established descriptions of gas behavior in coal. Methane is stored in coal largely in an adsorbed state on the walls of microscopic pores rather than as free gas in open voids, a property captured by Langmuir-type isotherms that relate adsorbed volume to gas pressure and temperature. As drainage lowers pressure near a borehole, adsorbed methane desorbs from the matrix, diffuses through the micropore network, and then flows as a fluid through the fracture system toward the well. Because coal is soft and compressible, the effective stress changes that accompany depressurization also deform the skeleton, opening or closing cleats and thereby altering permeability dynamically. Coupling these processes, as the COMSOL framework does, reflects the accepted understanding that coal seam permeability is not a static material property but an evolving quantity shaped by stress, pressure, and thermal conditions.

The choice of a hybrid machine learning architecture addresses a genuine methodological tension. Gradient-boosted tree ensembles such as XGBoost excel on tabular data and small sample counts, where their regularization and ensembling limit overfitting, but they do not naturally encode sequential structure. Recurrent networks such as LSTM are built to retain temporal context across long input sequences, which suits cumulative extraction curves that evolve over months of drainage, yet they typically demand far more training data than were available from twelve simulated schemes. Routing information from an LSTM encoder into a gradient-boosting regressor is a pragmatic compromise, letting the recurrent layer summarize temporal patterns while the tree ensemble handles the final nonlinear mapping with limited samples, a strategy consistent with broader trends in geoscience where physics simulations substitute for scarce field measurements.

The hyperparameter tuning step also deserves note. Models like XGBoost and LSTM are sensitive to settings such as learning rate, tree depth, and the number of hidden units, and manual tuning rarely explores this space efficiently. Metaheuristic optimizers inspired by animal foraging and flocking behavior offer a population-based alternative that can escape poor local optima without gradient information. Comparing four such optimizers, as the authors did, acknowledges that no single search strategy dominates every problem, and their convergence behavior on error surfaces derived from simulation data provides a useful empirical record for practitioners facing similar tuning burdens.

The finding that permeability dominates prediction aligns with decades of coalbed methane experience. Chinese coal seams, particularly in tectonically deformed regions such as Guizhou, are frequently characterized as low-permeability, high-gas-content reservoirs, which is why the engineering literature devotes such attention to stimulation. Hydraulic fracturing, hydraulic slotting, protective blasting, liquid nitrogen freeze-thaw cycling, and other enhancement techniques all aim to widen the fracture network through which desorbed gas can travel. A predictive tool that quantifies how strongly permeability improvements translate into recovered gas volume gives engineers a quantitative basis for deciding where stimulation investment yields the greatest return, rather than relying on empirical rules of thumb alone.

The Entropy Weight Method used for model comparison offers a reproducible alternative to subjective weighting of error metrics. Because MAE, MAPE, MSE, RMSE, and the coefficient of determination emphasize different aspects of performance, absolute error versus relative error versus squared-error sensitivity to large deviations, aggregating them requires defensible weights. Entropy weighting derives those weights from the information content of each metric across candidate models, penalizing metrics that fail to discriminate among models and rewarding those that do.

More broadly, the study exemplifies a growing paradigm in which numerical simulation generates training data for machine learning surrogates. Once trained, such surrogates can evaluate thousands of drainage scenarios in seconds that would each take hours or days to simulate physically, enabling optimization loops, uncertainty analysis, and eventually real-time decision support at the working face. As field validation accumulates across diverse geological settings, this hybrid simulation-plus-learning workflow could become standard infrastructure for intelligent, low-emission coal mining.

Subject of Research: Machine learning prediction of coal rock gas extraction volume supported by multi-physics numerical simulation for low-carbon coal mining

Article Title: Low carbon advancement through cleaner production: gas extraction simulation and machine learning model prediction of coal rock gas volume

Article References: Cai, J., Li, X., & Chen, S. (2026). Low carbon advancement through cleaner production: gas extraction simulation and machine learning model prediction of coal rock gas volume. Journal of Saudi Chemical Society, 30(5), Article 66. https://doi.org/10.1007/s44442-026-00119-0

Image Credits: AI Generated

DOI: 10.1007/s44442-026-00119-0

Keywords: coalbed methane, gas extraction, machine learning, XGBoost, LSTM, COMSOL simulation, optimization algorithms, coal seam permeability, carbon emissions, mine safety, predictive modeling, cleaner production