Pub Date : 2026-03-17DOI: 10.1016/j.psep.2026.108745
Xue Xingzhuo, Liu Yongcheng, Zhang Guowei, Xu Xue, Liu Lili, Zhang Shiyong, He Lu, Liu Chunyuan
Thermal runaway (TR) and propagation in lithium-ion battery (LIB) constitute one of the primary risks leading to fire and explosion accidents in energy storage system (ESS), highlighting the urgent need for efficient, environmentally friendly, and cost-effective suppression technologies. In this study, a 50 Ah LIB is investigated, and a dry ice jet suppression system using CO2 as the propellant gas is developed. The suppression effects of different dry ice injection duration (30s, 60s, 90s) on single-cell TR and TRP within a battery module are systematically examined. The results show that the 50 Ah battery releases approximately 177.8 kJ-291.8kJ of self-generated heat during TR, accompanied by the ejection of flammable gases carrying 80.3 kJ-117.9kJ of heat. Dry ice effectively suppresses TR through rapid sublimation and heat absorption. The 30s injection duration can quickly extinguish flames but is insufficient to interrupt the internal reaction; when the injection duration is extended to 60s and 90s, the dry ice absorbs 148.7kJ and 238.0kJ of heat from the single cell, respectively, effectively suppressing TR. For the battery module, dry ice effectively prevents TRP at injection duration of 60s and above, reducing the maximum temperature of adjacent batteries by 53% and 83%, respectively, compared with the thermal runaway trigger temperature of control group. Based on a combined analysis of cooling efficiency and heat absorption per unit mass, the 90s dry ice injection duration provides the best suppression performance. These results provide quantitative experimental evidence supporting the engineering application of dry ice in the thermal runaway protection of LIB in EES.
{"title":"Experimental study on the effectiveness of dry ice in suppressing thermal runaway and its propagation in 50 Ah lithium-ion batteries","authors":"Xue Xingzhuo, Liu Yongcheng, Zhang Guowei, Xu Xue, Liu Lili, Zhang Shiyong, He Lu, Liu Chunyuan","doi":"10.1016/j.psep.2026.108745","DOIUrl":"https://doi.org/10.1016/j.psep.2026.108745","url":null,"abstract":"Thermal runaway (TR) and propagation in lithium-ion battery (LIB) constitute one of the primary risks leading to fire and explosion accidents in energy storage system (ESS), highlighting the urgent need for efficient, environmentally friendly, and cost-effective suppression technologies. In this study, a 50 Ah LIB is investigated, and a dry ice jet suppression system using CO<ce:inf loc=\"post\">2</ce:inf> as the propellant gas is developed. The suppression effects of different dry ice injection duration (30<ce:hsp sp=\"0.25\"></ce:hsp>s, 60<ce:hsp sp=\"0.25\"></ce:hsp>s, 90<ce:hsp sp=\"0.25\"></ce:hsp>s) on single-cell TR and TRP within a battery module are systematically examined. The results show that the 50 Ah battery releases approximately 177.8 kJ-291.8<ce:hsp sp=\"0.25\"></ce:hsp>kJ of self-generated heat during TR, accompanied by the ejection of flammable gases carrying 80.3 kJ-117.9<ce:hsp sp=\"0.25\"></ce:hsp>kJ of heat. Dry ice effectively suppresses TR through rapid sublimation and heat absorption. The 30<ce:hsp sp=\"0.25\"></ce:hsp>s injection duration can quickly extinguish flames but is insufficient to interrupt the internal reaction; when the injection duration is extended to 60<ce:hsp sp=\"0.25\"></ce:hsp>s and 90<ce:hsp sp=\"0.25\"></ce:hsp>s, the dry ice absorbs 148.7<ce:hsp sp=\"0.25\"></ce:hsp>kJ and 238.0<ce:hsp sp=\"0.25\"></ce:hsp>kJ of heat from the single cell, respectively, effectively suppressing TR. For the battery module, dry ice effectively prevents TRP at injection duration of 60<ce:hsp sp=\"0.25\"></ce:hsp>s and above, reducing the maximum temperature of adjacent batteries by 53% and 83%, respectively, compared with the thermal runaway trigger temperature of control group. Based on a combined analysis of cooling efficiency and heat absorption per unit mass, the 90<ce:hsp sp=\"0.25\"></ce:hsp>s dry ice injection duration provides the best suppression performance. These results provide quantitative experimental evidence supporting the engineering application of dry ice in the thermal runaway protection of LIB in EES.","PeriodicalId":20743,"journal":{"name":"Process Safety and Environmental Protection","volume":"130 1","pages":""},"PeriodicalIF":7.8,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465885","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The catalytic conversion of CO2 to formic acid (FA) represents a promising route for carbon utilization and sustainable chemical production. However, this process faces thermodynamic limitations and requires highly efficient catalysts. Conventional synthesis methods often lead to structural inhomogeneity and involve complex procedures, constraining their efficiency and scalability. To overcome these limitations, we developed a one-pot solvothermal approach combined with calcination to fabricate Ru/CeO2 catalysts. The as-synthesized catalysts feature atomically dispersed Ru species, a significantly increased concentration of oxygen vacancies on CeO2, and enhanced metal-support interactions, which work synergistically to promote reactant activation and catalytic performance. Under mild reaction conditions, the optimized Ru/CeO2 catalyst delivered high performance, achieving a turnover number (TON) of 1112 and a turnover frequency (TOF) of 139h-1. Moreover, the catalyst maintained high stability and recyclability over multiple reaction cycles, underscoring its potential for industrial implementation. This study offers an efficient and scalable synthesis strategy for designing high-performance heterogeneous catalysts, thereby advancing carbon emission mitigation and sustainable chemical synthesis.
{"title":"One-Pot Synthesis of Single-Atom Ru-Immobilized Hierarchical Porous CeO2 Catalysts for CO2 Hydrogenation","authors":"Zhenzhen Wang, Liujun Wang, Chen Yuan, Bengao Yuan, Zhe Wang, Zhonghua Sun, Zhihui Zhang, Bing Lu, Ping Liu, Mingyang He, Junfeng Qian","doi":"10.1016/j.psep.2026.108744","DOIUrl":"https://doi.org/10.1016/j.psep.2026.108744","url":null,"abstract":"The catalytic conversion of CO<ce:inf loc=\"post\">2</ce:inf> to formic acid (FA) represents a promising route for carbon utilization and sustainable chemical production. However, this process faces thermodynamic limitations and requires highly efficient catalysts. Conventional synthesis methods often lead to structural inhomogeneity and involve complex procedures, constraining their efficiency and scalability. To overcome these limitations, we developed a one-pot solvothermal approach combined with calcination to fabricate Ru/CeO<ce:inf loc=\"post\">2</ce:inf> catalysts. The as-synthesized catalysts feature atomically dispersed Ru species, a significantly increased concentration of oxygen vacancies on CeO<ce:inf loc=\"post\">2</ce:inf>, and enhanced metal-support interactions, which work synergistically to promote reactant activation and catalytic performance. Under mild reaction conditions, the optimized Ru/CeO<ce:inf loc=\"post\">2</ce:inf> catalyst delivered high performance, achieving a turnover number (TON) of 1112 and a turnover frequency (TOF) of 139<ce:hsp sp=\"0.25\"></ce:hsp>h<ce:sup loc=\"post\">-1</ce:sup>. Moreover, the catalyst maintained high stability and recyclability over multiple reaction cycles, underscoring its potential for industrial implementation. This study offers an efficient and scalable synthesis strategy for designing high-performance heterogeneous catalysts, thereby advancing carbon emission mitigation and sustainable chemical synthesis.","PeriodicalId":20743,"journal":{"name":"Process Safety and Environmental Protection","volume":"52 1","pages":""},"PeriodicalIF":7.8,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465882","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Rock wettability is a critical parameter controlling residual trapping. However, wettability effects reported in literature appear inconsistent, and systematic studies quantifying influence of wettability alteration on CO2 migration and residual trapping remain limited. In this study, a series of drainage and imbibition NMR-based experiments were conducted under different wettability states (strongly water-wet, weakly water-wet, and intermediate-wet). Results indicated that enhanced water-wetness altered displacement patterns: the displacement uniformity index (DUI, the ratio of brine saturation changes during early-stage to late-stage drainage) increased from 2.36 to 3.96 with increasing capillary number (Ca), and magnetic resonance imaging (MRI) revealed the evolution of displacement front from capillary fingering to viscous fingering, resulting in reduced initial CO2 saturation. Pore-scale analysis revealed that CO2 migration into smallpores was suppressed under strongly water-wet, with negative convection phenomena occurring in micropores. The migration differences were attributed to higher capillary pressure and narrower two-phase effective permeability range under strongly water-wet. Notably, compared to intermediate-wet, strongly water-wet samples achieved lower initial CO2 saturation but higher residual CO2 saturation, attributed to enhanced capillary force that promoted CO2 snap-off. These findings highlight the potential of enhanced water-wetness for improving residual trapping, but lower initial CO2 saturation means more pore volume is needed to store a given amount of CO2, and CO2 injectivity may also be adversely affected. Moreover, Spiteri model analysis based on a literature database indicated that wettability effects on residual trapping are also influenced by rock type, permeability, and spatial wettability distribution, which should be considered in practical applications.
{"title":"Effect of wettability alteration on CO2 migration and residual trapping in saline aquifers: NMR-based experimental study","authors":"Jia Zhao, Chuanjin Yao, Yuyuan Song, Zhicheng Liu, Xingheng Huang","doi":"10.1016/j.psep.2026.108742","DOIUrl":"https://doi.org/10.1016/j.psep.2026.108742","url":null,"abstract":"Rock wettability is a critical parameter controlling residual trapping. However, wettability effects reported in literature appear inconsistent, and systematic studies quantifying influence of wettability alteration on CO<ce:inf loc=\"post\">2</ce:inf> migration and residual trapping remain limited. In this study, a series of drainage and imbibition NMR-based experiments were conducted under different wettability states (strongly water-wet, weakly water-wet, and intermediate-wet). Results indicated that enhanced water-wetness altered displacement patterns: the displacement uniformity index (DUI, the ratio of brine saturation changes during early-stage to late-stage drainage) increased from 2.36 to 3.96 with increasing capillary number (<ce:italic>Ca</ce:italic>), and magnetic resonance imaging (MRI) revealed the evolution of displacement front from capillary fingering to viscous fingering, resulting in reduced initial CO<ce:inf loc=\"post\">2</ce:inf> saturation. Pore-scale analysis revealed that CO<ce:inf loc=\"post\">2</ce:inf> migration into smallpores was suppressed under strongly water-wet, with negative convection phenomena occurring in micropores. The migration differences were attributed to higher capillary pressure and narrower two-phase effective permeability range under strongly water-wet. Notably, compared to intermediate-wet, strongly water-wet samples achieved lower initial CO<ce:inf loc=\"post\">2</ce:inf> saturation but higher residual CO<ce:inf loc=\"post\">2</ce:inf> saturation, attributed to enhanced capillary force that promoted CO<ce:inf loc=\"post\">2</ce:inf> snap-off. These findings highlight the potential of enhanced water-wetness for improving residual trapping, but lower initial CO<ce:inf loc=\"post\">2</ce:inf> saturation means more pore volume is needed to store a given amount of CO<ce:inf loc=\"post\">2</ce:inf>, and CO<ce:inf loc=\"post\">2</ce:inf> injectivity may also be adversely affected. Moreover, Spiteri model analysis based on a literature database indicated that wettability effects on residual trapping are also influenced by rock type, permeability, and spatial wettability distribution, which should be considered in practical applications.","PeriodicalId":20743,"journal":{"name":"Process Safety and Environmental Protection","volume":"57 1","pages":""},"PeriodicalIF":7.8,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465883","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-16DOI: 10.1016/j.psep.2026.108739
Liang Liu, Li Mo, Shuai Zhao, Mo He, Yong Hao, Chao Chen
Oil and gas pipeline defects such as corrosion may induce the release of toxic and flammable substances, resulting in personnel poisoning, fires, explosions, environmental pollution, and economic losses. Magnetic flux leakage (MFL) testing safeguards the integrity of oil and gas pipelines by enabling early identification and quantification of defects and thus avoiding accidents. This study establishes an experimental MFL measurement platform and develops a complete signal conditioning and data processing workflow to provide reliable inputs for defect sizing. Raw MFL signals collected from 315 artificial defects are standardized via unified window extraction, denoised using optimally selected wavelet decomposition parameters, and further augmented through random temporal shifting and time-sequence reversal to improve robustness against temporal disturbances. Based on the processed axial and radial MFL data, a dual-channel convolutional neural network–bidirectional long short-term memory (DC-CNN-BiLSTM) model is proposed for three-dimensional defect quantification. The dual-channel convolutional structure captures complementary local and global spatial characteristics through parallel standard and dilated convolutions, while the BiLSTM module models temporal dependencies inherent to magnetic leakage patterns. Experimental results demonstrate high measurement accuracy, achieving prediction accuracies of 95.8% (length), 92.7% (width), and 93.7% (depth), with RMSE values of 1.606mm, 4.080mm, and 0.387mm, respectively. Ablation studies confirm the effectiveness of both dual-channel feature extraction and bidirectional sequence modeling, while data augmentation improves overall accuracy by 3.4%. These findings indicate that integrating high-quality MFL data with hybrid deep learning provides an effective and practical solution for accurate defect quantification and supports more reliable decision-making in pipeline integrity and process safety management.
{"title":"Experimental data-driven defect quantification of oil and gas pipelines using magnetic flux leakage inspection: A hybrid neural network for safety risk identification","authors":"Liang Liu, Li Mo, Shuai Zhao, Mo He, Yong Hao, Chao Chen","doi":"10.1016/j.psep.2026.108739","DOIUrl":"https://doi.org/10.1016/j.psep.2026.108739","url":null,"abstract":"Oil and gas pipeline defects such as corrosion may induce the release of toxic and flammable substances, resulting in personnel poisoning, fires, explosions, environmental pollution, and economic losses. Magnetic flux leakage (MFL) testing safeguards the integrity of oil and gas pipelines by enabling early identification and quantification of defects and thus avoiding accidents. This study establishes an experimental MFL measurement platform and develops a complete signal conditioning and data processing workflow to provide reliable inputs for defect sizing. Raw MFL signals collected from 315 artificial defects are standardized via unified window extraction, denoised using optimally selected wavelet decomposition parameters, and further augmented through random temporal shifting and time-sequence reversal to improve robustness against temporal disturbances. Based on the processed axial and radial MFL data, a dual-channel convolutional neural network–bidirectional long short-term memory (DC-CNN-BiLSTM) model is proposed for three-dimensional defect quantification. The dual-channel convolutional structure captures complementary local and global spatial characteristics through parallel standard and dilated convolutions, while the BiLSTM module models temporal dependencies inherent to magnetic leakage patterns. Experimental results demonstrate high measurement accuracy, achieving prediction accuracies of 95.8% (length), 92.7% (width), and 93.7% (depth), with RMSE values of 1.606<ce:hsp sp=\"0.25\"></ce:hsp>mm, 4.080<ce:hsp sp=\"0.25\"></ce:hsp>mm, and 0.387<ce:hsp sp=\"0.25\"></ce:hsp>mm, respectively. Ablation studies confirm the effectiveness of both dual-channel feature extraction and bidirectional sequence modeling, while data augmentation improves overall accuracy by 3.4%. These findings indicate that integrating high-quality MFL data with hybrid deep learning provides an effective and practical solution for accurate defect quantification and supports more reliable decision-making in pipeline integrity and process safety management.","PeriodicalId":20743,"journal":{"name":"Process Safety and Environmental Protection","volume":"87 1","pages":""},"PeriodicalIF":7.8,"publicationDate":"2026-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465886","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-16DOI: 10.1016/j.psep.2026.108738
Yue Li, Tao Yang, Dongmei Fu
Corrosion prediction modeling for metallic materials is a classic challenge in materials research. However, it has long been constrained by the multitude of influencing factors and their complex coupling relationships, making it a challenging area of study that has attracted widespread attention. To comprehensively address the interrelated factors in material corrosion, this study adopts an advanced deep learning approach. It employs feature analysis and fusion learning of multi-source heterogeneous data to uncover the patterns governing material corrosion evolution. To this end, we propose a Deep learning-based Multi-source data fusion model for Corrosion rate prediction (DMCorr). DMCorr integrates material composition, meteorological time series data, and macro-corrosion images, to predict material corrosion rates under specific environmental conditions. This model features three distinct modules that extract key features from each data source, respectively. In particular, for meteorological time-series data, we introduce time series decomposition processing, enabling the model to better capture the deep features of dynamic meteorological factors. It further incorporates a Feature-wise Linear Modulation (FiLM) network to dynamically and adaptively fuse these features. In addition, the overall framework adopts a two-stage training strategy: first, the unimodal encoders for the time-series and image data are pretrained separately and then frozen; subsequently, the material encoder, the FiLM-based fusion module, and the prediction head are jointly optimized. This design improves the stability and training efficiency of multi-source fusion under limited-data conditions. The performance of the proposed model was compared against benchmark models using the coupons exposure data from multiple sites. The results show that, compared with the best-performing baseline model, DMCorr reduces the MAE by 32.59% and the RMSE by 12.06%, while improving the coefficient of determination R2 to 0.8749, thereby demonstrating the effectiveness and feasibility of the proposed model.
{"title":"DMCorr: A Deep learning-based Multi-source data fusion model for Corrosion rate prediction of metallic materials","authors":"Yue Li, Tao Yang, Dongmei Fu","doi":"10.1016/j.psep.2026.108738","DOIUrl":"https://doi.org/10.1016/j.psep.2026.108738","url":null,"abstract":"Corrosion prediction modeling for metallic materials is a classic challenge in materials research. However, it has long been constrained by the multitude of influencing factors and their complex coupling relationships, making it a challenging area of study that has attracted widespread attention. To comprehensively address the interrelated factors in material corrosion, this study adopts an advanced deep learning approach. It employs feature analysis and fusion learning of multi-source heterogeneous data to uncover the patterns governing material corrosion evolution. To this end, we propose a Deep learning-based Multi-source data fusion model for Corrosion rate prediction (DMCorr). DMCorr integrates material composition, meteorological time series data, and macro-corrosion images, to predict material corrosion rates under specific environmental conditions. This model features three distinct modules that extract key features from each data source, respectively. In particular, for meteorological time-series data, we introduce time series decomposition processing, enabling the model to better capture the deep features of dynamic meteorological factors. It further incorporates a Feature-wise Linear Modulation (FiLM) network to dynamically and adaptively fuse these features. In addition, the overall framework adopts a two-stage training strategy: first, the unimodal encoders for the time-series and image data are pretrained separately and then frozen; subsequently, the material encoder, the FiLM-based fusion module, and the prediction head are jointly optimized. This design improves the stability and training efficiency of multi-source fusion under limited-data conditions. The performance of the proposed model was compared against benchmark models using the coupons exposure data from multiple sites. The results show that, compared with the best-performing baseline model, DMCorr reduces the MAE by 32.59% and the RMSE by 12.06%, while improving the coefficient of determination <mml:math altimg=\"si180.svg\" display=\"inline\"><mml:msup><mml:mrow><mml:mi>R</mml:mi></mml:mrow><mml:mrow><mml:mn>2</mml:mn></mml:mrow></mml:msup></mml:math> to 0.8749, thereby demonstrating the effectiveness and feasibility of the proposed model.","PeriodicalId":20743,"journal":{"name":"Process Safety and Environmental Protection","volume":"282 1","pages":""},"PeriodicalIF":7.8,"publicationDate":"2026-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465887","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-16DOI: 10.1016/j.psep.2026.108737
K. Golobish, Z. Chua, C. Cooper
Pipes are fundamental to resource distribution and play an important role in residential and industrial infrastructure. Therefore, accurate detection and timely remediation of pipe leaks are critical to avoiding structural damage and environmental contamination. Although effective leak detection methods have been recently developed that employ neural networks to analyze wavelet-decomposed pipe vibrations, these techniques require large training datasets unique to each pipe system being analyzed, the collection of which is often infeasible due to practical, financial, and safety constraints. This drawback is compounded by the sensitivity of the continuous wavelet transform, which is commonly employed for feature extraction, to the choice of mother wavelet. These two factors combine such that existing leak detection approaches perform poorly on out-of-distribution data. To address these limitations, we propose augmenting ground-truth training data with synthetic pipe leak vibration signals sampled from a Gaussian Mixture Model. The mixture model is trained to learn a low-dimensional frequency-domain representation of the vibration signals from which novel signals are sampled to generate time-domain vibration data compatible with a myriad of mother wavelets. Experimental validation showed that incorporating synthetic data into the training data increased leak detection recall by 10.0 pp. on average for samples outside of the ground truth vibration data distribution, with all wavelet choices achieving a minimum of 99% leak detection accuracy and 99% recall. A significant decrease in classification loss was also observed for sub-optimal wavelet choices as the proportion of synthetic data in the training dataset increased.
{"title":"Enhanced pipe leak detection via Gaussian mixture model-based data synthesis and wavelet-agnostic feature extraction","authors":"K. Golobish, Z. Chua, C. Cooper","doi":"10.1016/j.psep.2026.108737","DOIUrl":"https://doi.org/10.1016/j.psep.2026.108737","url":null,"abstract":"Pipes are fundamental to resource distribution and play an important role in residential and industrial infrastructure. Therefore, accurate detection and timely remediation of pipe leaks are critical to avoiding structural damage and environmental contamination. Although effective leak detection methods have been recently developed that employ neural networks to analyze wavelet-decomposed pipe vibrations, these techniques require large training datasets unique to each pipe system being analyzed, the collection of which is often infeasible due to practical, financial, and safety constraints. This drawback is compounded by the sensitivity of the continuous wavelet transform, which is commonly employed for feature extraction, to the choice of mother wavelet. These two factors combine such that existing leak detection approaches perform poorly on out-of-distribution data. To address these limitations, we propose augmenting ground-truth training data with synthetic pipe leak vibration signals sampled from a Gaussian Mixture Model. The mixture model is trained to learn a low-dimensional frequency-domain representation of the vibration signals from which novel signals are sampled to generate time-domain vibration data compatible with a myriad of mother wavelets. Experimental validation showed that incorporating synthetic data into the training data increased leak detection recall by 10.0 pp. on average for samples outside of the ground truth vibration data distribution, with all wavelet choices achieving a minimum of 99% leak detection accuracy and 99% recall. A significant decrease in classification loss was also observed for sub-optimal wavelet choices as the proportion of synthetic data in the training dataset increased.","PeriodicalId":20743,"journal":{"name":"Process Safety and Environmental Protection","volume":"160 1","pages":""},"PeriodicalIF":7.8,"publicationDate":"2026-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465891","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-16DOI: 10.1016/j.psep.2026.108741
Menghui Zhang, Zonghou Huang, Qingsong Wang, Fuqiang Yang
Thermal runaway (TR) in power banks poses serious safety concerns due to their widespread use and compact enclosed structures. This study experimentally investigates the TR behavior and early warning characteristics of power bank cells (PBCs) and power bank units (PBUs) under localized overheating abuse at different states of charge (SOCs) for the first time. Multidimensional signals, including temperature, voltage, expansion force, and mass loss, are synchronously monitored, and infrared thermography is employed to characterize surface temperature evolution. Results indicate that the casing partially mitigates TR severity by elevating the trigger temperature and suppressing flames at 75%SOC, yet simultaneously randomizes jet flame and smoke discharge paths, potentially increasing fire risk. Expansion force and casing surface temperature are identified as effective early warning indicators. Abnormal front casing temperature rise (dT/dt ≥ 1 ℃/s) and expansion force increase (dF/dt ≥ 0.5N/ s) provide earlier warnings than voltage signals, corresponding to 529-730s and 489-680s prior to TR onset. Based on these findings, a four-level early warning framework for PBU is proposed for hierarchical risk identification under realistic operating conditions. This work offers a practical and implementable strategy for early TR risk identification and safety design of enclosed lithium-ion power bank systems.
{"title":"Thermal runaway behavior and multi-signal early warning strategies for lithium-ion power banks under localized overheating abuse","authors":"Menghui Zhang, Zonghou Huang, Qingsong Wang, Fuqiang Yang","doi":"10.1016/j.psep.2026.108741","DOIUrl":"https://doi.org/10.1016/j.psep.2026.108741","url":null,"abstract":"Thermal runaway (TR) in power banks poses serious safety concerns due to their widespread use and compact enclosed structures. This study experimentally investigates the TR behavior and early warning characteristics of power bank cells (PBCs) and power bank units (PBUs) under localized overheating abuse at different states of charge (SOCs) for the first time. Multidimensional signals, including temperature, voltage, expansion force, and mass loss, are synchronously monitored, and infrared thermography is employed to characterize surface temperature evolution. Results indicate that the casing partially mitigates TR severity by elevating the trigger temperature and suppressing flames at 75%SOC, yet simultaneously randomizes jet flame and smoke discharge paths, potentially increasing fire risk. Expansion force and casing surface temperature are identified as effective early warning indicators. Abnormal front casing temperature rise (d<ce:italic>T</ce:italic>/d<ce:italic>t</ce:italic> ≥ 1 ℃/s) and expansion force increase (d<ce:italic>F</ce:italic>/d<ce:italic>t</ce:italic> ≥ 0.5<ce:hsp sp=\"0.25\"></ce:hsp>N/ s) provide earlier warnings than voltage signals, corresponding to 529-730<ce:hsp sp=\"0.25\"></ce:hsp>s and 489-680<ce:hsp sp=\"0.25\"></ce:hsp>s prior to TR onset. Based on these findings, a four-level early warning framework for PBU is proposed for hierarchical risk identification under realistic operating conditions. This work offers a practical and implementable strategy for early TR risk identification and safety design of enclosed lithium-ion power bank systems.","PeriodicalId":20743,"journal":{"name":"Process Safety and Environmental Protection","volume":"12 1","pages":""},"PeriodicalIF":7.8,"publicationDate":"2026-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465884","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-16DOI: 10.1016/j.psep.2026.108736
Bohan Yao, Xiaogang Deng, Ping Wang
Deep support vector data description (Deep SVDD) has emerged as a popular chemical process fault detection technology. However, traditional deep SVDD method focuses on the static relationship mining, but omits the dynamic fault evolving characteristic (i.e., fault symptoms shift across variables and time). Moreover, as an unsupervised one-class modeling method, deep SVDD fails to utilize prior fault knowledge for boundary optimization, resulting in insensitive fault detection. To address these problems, this paper proposes an enhanced deep SVDD method, called Probe-Recall Deep SVDD (PR-DSVDD), for more effective chemical process fault detection. In the proposed method, two novel modules of representation probe and memory recall are designed for evolving characteristic representation and decision boundary refining, respectively. In the representation probe module, a dynamic spatiotemporal evolution encoder is developed to facilitate adaptive attention mechanism. This module captures the evolving weight distribution of critical channels across time steps, effectively modeling the symptom shifts and position-sensitive patterns in fault propagation. The memory recall module constructs a contrastive memory mechanism that leverages a few fault prototype features to actively calibrate the decision boundary. Further, a fault evolution explanation diagram based on relative SHAP values and parallel coordinate system is presented to visualize the fault propagation procedure. Experiments on the Tennessee Eastman process demonstrate that PR-DSVDD is superior to the compared fault detection methods by achieving an average detection rate of 82.88% with a low false positive rate of 1.28%, and the proposed fault evolution explanation plot captures physically consistent fault propagation paths, offering physically consistent explanations.
{"title":"Evolving fault detection for chemical processes via probe-recall deep SVDD with adaptive attention and contrastive memory","authors":"Bohan Yao, Xiaogang Deng, Ping Wang","doi":"10.1016/j.psep.2026.108736","DOIUrl":"https://doi.org/10.1016/j.psep.2026.108736","url":null,"abstract":"Deep support vector data description (Deep SVDD) has emerged as a popular chemical process fault detection technology. However, traditional deep SVDD method focuses on the static relationship mining, but omits the dynamic fault evolving characteristic (i.e., fault symptoms shift across variables and time). Moreover, as an unsupervised one-class modeling method, deep SVDD fails to utilize prior fault knowledge for boundary optimization, resulting in insensitive fault detection. To address these problems, this paper proposes an enhanced deep SVDD method, called Probe-Recall Deep SVDD (PR-DSVDD), for more effective chemical process fault detection. In the proposed method, two novel modules of representation probe and memory recall are designed for evolving characteristic representation and decision boundary refining, respectively. In the representation probe module, a dynamic spatiotemporal evolution encoder is developed to facilitate adaptive attention mechanism. This module captures the evolving weight distribution of critical channels across time steps, effectively modeling the symptom shifts and position-sensitive patterns in fault propagation. The memory recall module constructs a contrastive memory mechanism that leverages a few fault prototype features to actively calibrate the decision boundary. Further, a fault evolution explanation diagram based on relative SHAP values and parallel coordinate system is presented to visualize the fault propagation procedure. Experiments on the Tennessee Eastman process demonstrate that PR-DSVDD is superior to the compared fault detection methods by achieving an average detection rate of 82.88% with a low false positive rate of 1.28%, and the proposed fault evolution explanation plot captures physically consistent fault propagation paths, offering physically consistent explanations.","PeriodicalId":20743,"journal":{"name":"Process Safety and Environmental Protection","volume":"11 1","pages":""},"PeriodicalIF":7.8,"publicationDate":"2026-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465890","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-15Epub Date: 2026-02-10DOI: 10.1016/j.psep.2026.108588
Ping Li , Yihui Jiang , Di Wu , Jian Guo , Jinwei Xu , Yiwen Xu , Bo Ke , Songlin Zhang , Qiangling Duan
With the rapid growth of hydrogen energy technologies, safety hazards associated with high-pressure hydrogen release in confined enclosures have become a critical concern. This study experimentally investigates hydrogen combustion and explosion induced by self-ignition flames following high-pressure hydrogen leakage. The effects of axial obstruction and ventilation conditions on flame dynamics, temperature evolution, and overpressure development were systematically examined. Results show that when the hydrogen jet is unobstructed, most hydrogen exits the enclosure and burns externally, while a symmetric flame forms along the central axis inside the enclosure. In contrast, axial obstruction generates complex turbulent flow, leading to spatially non-uniform hydrogen/air mixtures. The self-ignition flame first ignites the central mixture, and combustion propagates throughout the enclosure, producing higher peak temperatures, longer combustion durations, and pronounced overpressure oscillations, including initial negative pressures. Increasing the ventilation area under obstructed conditions enhances air entrainment, further intensifying combustion. These findings reveal that combustion mechanisms and associated hazards are fundamentally different between obstructed and unobstructed hydrogen releases. The study provides critical experimental evidence and mechanistic insights for the design of safe hydrogen storage, handling, and transportation systems in confined environments.
{"title":"Experimental investigation of hydrogen self-ignition and explosion in confined enclosures: Influence of axial obstruction and ventilation","authors":"Ping Li , Yihui Jiang , Di Wu , Jian Guo , Jinwei Xu , Yiwen Xu , Bo Ke , Songlin Zhang , Qiangling Duan","doi":"10.1016/j.psep.2026.108588","DOIUrl":"10.1016/j.psep.2026.108588","url":null,"abstract":"<div><div>With the rapid growth of hydrogen energy technologies, safety hazards associated with high-pressure hydrogen release in confined enclosures have become a critical concern. This study experimentally investigates hydrogen combustion and explosion induced by self-ignition flames following high-pressure hydrogen leakage. The effects of axial obstruction and ventilation conditions on flame dynamics, temperature evolution, and overpressure development were systematically examined. Results show that when the hydrogen jet is unobstructed, most hydrogen exits the enclosure and burns externally, while a symmetric flame forms along the central axis inside the enclosure. In contrast, axial obstruction generates complex turbulent flow, leading to spatially non-uniform hydrogen/air mixtures. The self-ignition flame first ignites the central mixture, and combustion propagates throughout the enclosure, producing higher peak temperatures, longer combustion durations, and pronounced overpressure oscillations, including initial negative pressures. Increasing the ventilation area under obstructed conditions enhances air entrainment, further intensifying combustion. These findings reveal that combustion mechanisms and associated hazards are fundamentally different between obstructed and unobstructed hydrogen releases. The study provides critical experimental evidence and mechanistic insights for the design of safe hydrogen storage, handling, and transportation systems in confined environments.</div></div>","PeriodicalId":20743,"journal":{"name":"Process Safety and Environmental Protection","volume":"209 ","pages":"Article 108588"},"PeriodicalIF":7.8,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146153054","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-15Epub Date: 2026-02-12DOI: 10.1016/j.psep.2026.108598
Haotian Pang , Haole Wang , Qian Tian , Hua Li , Zecong Zhou , Yujiang Wang
Solidification/stabilization is a key technology for efficiently immobilizing heavy metals and enabling the resourceful utilization of solid waste. In this study, a novel low-carbon cementitious material (LCM) was developed by using red mud (RM), calcium carbide slag (CS), and phosphogypsum (PG) as composite alkaline activators, while synergistically activating ground granulated blast furnace slag (GGBS) and fly ash (FA). The effects of different Cr3+ contents and material ratios on the system's mechanical properties, leaching characteristics, and microstructure were investigated, revealing a multi-pathway solidification mechanism of Cr3+. The results show that Cr3+ acts as a "regulator" within a certain range, rather than simply an inhibitor. In the early stages, Cr3+ reacts with OH− in the solution, weakening the early hydration of the matrix. However, over the long term, this delay actually facilitates the ordered diffusion and polymerization of [SiO4]4− and [AlO4]5− in the gel system, thereby enhancing the crosslinking of C−(A)−S−H and N−A−S−H gels. The study further shows that Cr3+ stabilizes and solidifies through three main pathways: substitution of Ca2+ in C−(A)−S−H gel, adsorption and encapsulation by the N−A−S−H gel, and replacement of Al3+ in the AFt crystal structure. Interestingly, when the CS content was maintained between 10 % and 15 %, a new phase, Ca−Cr layered double hydroxide, appeared, providing a new mechanism for Cr3+ fixation. The fixation rate of Cr3+ in all samples exceeds 99.97 %, and compressive strength reaches 30.8 MPa, an increase of 21.5 %, indicating excellent fixation effects and environmental compatibility of LCM with Cr3+.
{"title":"Development of low-carbon cementitious materials for Cr(III) immobilization: Mechanisms of solidification, stabilization, and structural enhancement","authors":"Haotian Pang , Haole Wang , Qian Tian , Hua Li , Zecong Zhou , Yujiang Wang","doi":"10.1016/j.psep.2026.108598","DOIUrl":"10.1016/j.psep.2026.108598","url":null,"abstract":"<div><div>Solidification/stabilization is a key technology for efficiently immobilizing heavy metals and enabling the resourceful utilization of solid waste. In this study, a novel low-carbon cementitious material (LCM) was developed by using red mud (RM), calcium carbide slag (CS), and phosphogypsum (PG) as composite alkaline activators, while synergistically activating ground granulated blast furnace slag (GGBS) and fly ash (FA). The effects of different Cr<sup>3</sup><sup>+</sup> contents and material ratios on the system's mechanical properties, leaching characteristics, and microstructure were investigated, revealing a multi-pathway solidification mechanism of Cr<sup>3+</sup>. The results show that Cr<sup>3+</sup> acts as a \"regulator\" within a certain range, rather than simply an inhibitor. In the early stages, Cr<sup>3+</sup> reacts with OH<sup>−</sup> in the solution, weakening the early hydration of the matrix. However, over the long term, this delay actually facilitates the ordered diffusion and polymerization of [SiO<sub>4</sub>]<sup>4−</sup> and [AlO<sub>4</sub>]<sup>5−</sup> in the gel system, thereby enhancing the crosslinking of C−(A)−S−H and N−A−S−H gels. The study further shows that Cr<sup>3+</sup> stabilizes and solidifies through three main pathways: substitution of Ca<sup>2+</sup> in C−(A)−S−H gel, adsorption and encapsulation by the N−A−S−H gel, and replacement of Al<sup>3+</sup> in the AFt crystal structure. Interestingly, when the CS content was maintained between 10 % and 15 %, a new phase, Ca−Cr layered double hydroxide, appeared, providing a new mechanism for Cr<sup>3+</sup> fixation. The fixation rate of Cr<sup>3+</sup> in all samples exceeds 99.97 %, and compressive strength reaches 30.8 MPa, an increase of 21.5 %, indicating excellent fixation effects and environmental compatibility of LCM with Cr<sup>3+</sup>.</div></div>","PeriodicalId":20743,"journal":{"name":"Process Safety and Environmental Protection","volume":"209 ","pages":"Article 108598"},"PeriodicalIF":7.8,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146172252","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：