Article
Computer Science, Artificial Intelligence
Junjie Shi, Jiang Bian, Jakob Richter, Kuan-Hsun Chen, Jorg Rahnenfuhrer, Haoyi Xiong, Jian-Jia Chen
Summary: The predictive performance of a machine learning model depends on hyper-parameter setting, making hyper-parameter tuning crucial. In distributed machine learning, collecting all data is challenging, thus the MODES framework is proposed to deploy MBO on resource-constrained distributed embedded systems to optimize combined prediction accuracy.
Article
Engineering, Civil
Roberto Falcone, Angelo Ciaramella, Francesco Carrabs, Nicola Strisciuglio, Enzo Martinelli
Summary: This paper proposes the use of Machine Learning as a substitute for traditional mechanistic analyses in seismic retrofitting of reinforced concrete structures. The obtained results demonstrate the effectiveness of an artificial neural network in rapidly and accurately assessing the performance of different structural configurations, and it can be used to speed up the search for the best retrofitting solution.
Article
Materials Science, Multidisciplinary
Jack G. Nedell, Jonah Spector, Adel Abbout, Michael Vogl, Gregory A. Fiete
Summary: Motivated by the improvement of deep learning techniques, we design a mixed input convolutional neural network to predict transport properties in deformed nanoscale materials using a height map of deformations as input. Our network achieves higher accuracy in conductance predictions by using redundant input of energy values, and it successfully predicts valley-resolved conductance.
Article
Chemistry, Multidisciplinary
Samaneh Rashtbari, Gholamreza Dehghan, Sirous Khorram, Mojtaba Amini, Alireza Khataee, Yeojoon Yoon
Summary: This study presents a simple and efficient approach for degrading triphenylmethane and malachite green using Argon cold plasma-modified cobalt oxide nanoparticles. The catalytic activity of the nanoparticles was enhanced after plasma modification, leading to the complete degradation of MG within 70 minutes. The produced metabolites were found to be less toxic, and a neural network model accurately predicted the efficiency of MG removal under different conditions.
JOURNAL OF INDUSTRIAL AND ENGINEERING CHEMISTRY
(2023)
Article
Engineering, Environmental
Tuhin Kamilya, Abhradeep Majumder, Duduku Saidulu, Subhasish Tripathy, Ashok K. Gupta
Summary: In this study, a continuous system combining a moving bed biofilm reactor (MBBR) with a horizontal subsurface flow constructed wetland (HSSFCW) was used for the removal of chemical oxygen demand (COD), ammonia (NH4+), and paracetamol. The system showed high removal efficiency for these pollutants by adjusting the initial concentration, hydraulic retention time (HRT), and COD to NH4+ ratio (C/N ratio). Further analysis revealed that microbial degradation, plant uptake, and substrate adsorption were the dominant removal mechanisms in the system.
CHEMICAL ENGINEERING JOURNAL
(2023)
Article
Engineering, Mechanical
Andreas Rosenkranz, Max Marian, Francisco J. Profito, Nathan Aragon, Raj Shah
Summary: This article introduces the application of artificial intelligence and machine learning in the field of tribology, highlighting recent advances and successful case studies using artificial neural networks. These methods demonstrate accurate and efficient prediction of tribological characteristics, with future research directions emphasizing the extended use of artificial intelligence and machine learning concepts.
Article
Energy & Fuels
Nawaf M. Alghamdi, S. Mani Sarathy
Summary: In this study, the partial and total oxidation of dimethyl ether (DME) over 5 wt% Rh/Al2O3 catalyst at low temperatures was investigated. The effects of temperature, flow rate, and inlet feed composition on the reactivity were studied. The experimental data provided valuable information for accurate kinetic modeling, reactor design and optimization, and rational catalyst design.
Article
Environmental Sciences
A. K. Maurya, B. S. Reddy, J. Theerthagiri, P. L. Narayana, C. H. Park, J. K. Hong, J-T Yeom, K. K. Cho, N. S. Reddy
Summary: This study explored the feasibility of using artificial neural networks to correlate biofilm reactor process parameters with absorption efficiency in handling industrial wastewater. The model predictions indicated that temperature and pH values are the most influential factors affecting absorption efficiency and turbidity.
SCIENCE OF THE TOTAL ENVIRONMENT
(2021)
Article
Energy & Fuels
R. Yukesh Kannah, K. Bhava Rohini, M. Gunasekaran, K. Gokulakrishnan, Gopalakrishnan Kumar, J. Rajesh Banu
Summary: In this study, a lab-scale reactor was used to treat landfill leachate, and an ANN model was developed to assess the substrate concentration and biogas yield. The results showed that an OLR of 16.27 kg COD/m(3)d was the optimum condition, resulting in the highest biogas production of 30.07 L/d with 64% methane content and 89.6% COD removal.
Article
Green & Sustainable Science & Technology
Komal Tripathi, Vrinda Gupta, Varsha Awasthi, Kamal Kishore Pant, Sreedevi Upadhyayula
Summary: This study develops an ultrafast machine learning (ML) based framework to predict CO2 conversion and methanol selectivity by extracting comprehensive knowledge from existing published literature. Among various ML algorithms, artificial neural networks (ANNs) exhibit the best accuracy. The efficacy and fidelity of the developed neural networks are depicted by satisfactory performance for majority of unseen test datasets. This work concludes that ML-based concept allows to uncover catalytic property-performance correlations hidden in existing experimental research.
ADVANCED SUSTAINABLE SYSTEMS
(2023)
Article
Chemistry, Multidisciplinary
Jinfu Lin, Hongxia Liu, Shulong Wang, Dong Wang, Lei Wu
Summary: This paper studies the hardware implementation of a fully connected neural network based on the 1T1R array and its application in handwritten digital image recognition. Experimental and simulation analysis show the relationship between the recognition accuracy of the network and the number of hidden neurons. The results indicate that the network has high recognition accuracy and the impact of failed devices on accuracy is minimal.
Article
Computer Science, Artificial Intelligence
Alejandro Moran, Vincent Canals, Fabio Galan-Prado, Christian F. Frasser, Dhinakar Radhakrishnan, Saeid Safavi, Josep L. Rossello
Summary: Edge artificial intelligence is a growing research field, and reservoir computing has attracted attention as a feasible alternative for edge intelligence. This study proposes a simple hardware-optimized circuit design for low-power edge intelligence applications and demonstrates its implementation in FPGA for low-power audio event detection. The results show significant accuracy and ultra-low energy consumption for the proposed approach.
COGNITIVE COMPUTATION
(2023)
Article
Chemistry, Physical
Faezeh Rahmani, Nima Ghal-Eh, Sergey Bedenko
Summary: The study focused on the feasibility of a landmine identification system (LIS) using the angular distribution of thermal neutrons, utilizing artificial neural networks and least-squares methods with input data prepared by the MCNP6.1 Monte Carlo code. Achieving promising results with a relative error of less than 15%, the study confirmed the system's sensitivity to landmine depth and soil moisture, suggesting its potential for landmine identification.
RADIATION PHYSICS AND CHEMISTRY
(2022)
Review
Chemistry, Multidisciplinary
Sungmin Han, C. Buddie Mullins
Summary: The study of Pd-Au model catalysts under UHV conditions reveals the influence of Pd ensemble size on H-2 desorption behavior, allowing H-2 to be used as a probe molecule for quantifying surface composition. Additionally, the Pd-Au interface is identified as the main reaction site for generating H-2, providing insights for potential liquid storage mediums for hydrogen.
ACCOUNTS OF CHEMICAL RESEARCH
(2021)
Article
Multidisciplinary Sciences
Xiangyan Meng, Guojie Zhang, Nuannuan Shi, Guangyi Li, Jose Azana, Jose Capmany, Jianping Yao, Yichen Shen, Wei Li, Ninghua Zhu, Ming Li
Summary: Convolutional neural networks are facing limitations in processing massive data due to electrical frequency and memory access time. Optical computing offers faster processing speeds and higher energy efficiency, but current schemes lack scalability. In this study, a compact on-chip optical convolutional processing unit is demonstrated on a low-loss silicon nitride platform, showing potential for large-scale integration.
NATURE COMMUNICATIONS
(2023)
Article
Engineering, Environmental
Napameth Phantawesak, Finn Coyle, Marc E. J. Stettler
Summary: This study presents real-world NOx emissions from 97 diesel-hybrid buses in London, showing that selective catalytic reduction (SCR) retrofitting can significantly reduce emissions. The study also indicates the impact of ambient temperature on NOx emissions, and highlights the effects of reduced congestion during the COVID-19 pandemic on SCR performance.
ENVIRONMENTAL SCIENCE & TECHNOLOGY
(2022)
Article
Computer Science, Interdisciplinary Applications
Mino Woo, Terry Jordan, Tarak Nandi, Jean Francois Dietiker, Christopher Guenther, Dirk Van Essendelft
Summary: Manufacturers are developing new GPU nodes that have large capacity, high bandwidth memory, and very high bandwidth intra-node interconnects. This allows for low-cost data movement between GPUs on the same node. However, the expensive global dot products due to small packet bandwidths and latencies can be mitigated by using equation decomposition instead of traditional domain decomposition. Testing this theory, the code MFiX was ported to TensorFlow and resulted in the accelerated MFiX-AI, which showed competitive performance to a supercomputer with 1000 CPU cores, achieving significant energy savings.
ENGINEERING WITH COMPUTERS
(2023)
Article
Environmental Studies
Swapnil S. Jagtap, Peter R. N. Childs, Marc E. J. Stettler
Summary: Decarbonising long-range aviation is challenging. This study evaluates the performance of six low-carbon fuels and their realistic impacts on aircraft design for a large long-range passenger aircraft using Breguet's range equation. Liquid hydrogen (LH2) and 100 % synthetic paraffin kerosene (SPK) are the only two alternative fuels found to be viable. Our results should inform studies on LH2 and 100 % SPK aircraft operating costs and lifecycle emissions.
TRANSPORTATION RESEARCH PART D-TRANSPORT AND ENVIRONMENT
(2023)
Article
Medicine, General & Internal
Kuan-Yuan Chen, Hsiao-Yun Kuo, Kang-Yun Lee, Po-Hao Feng, Sheng-Ming Wu, Hsiao-Chi Chuang, Tzu-Tao Chen, Wei-Lun Sun, Chien-Hua Tseng, Wen-Te Liu, Wun-Hao Cheng, Arnab Majumdar, Marc Stettler, Cheng-Yu Tsai, Shu-Chuan Ho
Summary: This study aimed to explore the feasibility of using indicators to evaluate the stability of COPD and their associations with other COPD-related clinical parameters. The results showed that a lower value of the distance-saturation product (DSP) was associated with greater worsening of symptoms, increased frequency of acute exacerbations, decreased pulmonary function, and more severe emphysema (higher low-attenuation area percentage, LAA%). Therefore, the DSP and LAA% can serve as indicators for assessing the clinical course of COPD and guiding corresponding treatments.
FRONTIERS IN MEDICINE
(2023)
Review
Energy & Fuels
Marc E. J. Stettler, Mino Woo, Daniel Ainalis, Pablo Achurra-Gonzalez, Jamie Speirs, Jasmin Cooper, Dong-Ha Lim, Nigel Brandon, Adam Hawkes
Summary: It has been suggested that using liquefied natural gas as a fuel source for heavy goods vehicles could lead to reduced greenhouse gas emissions. However, there has been little comparative analysis across various studies. This review provides a comprehensive examination of the well-to-wheel lifecycle emissions of liquefied natural gas for heavy goods vehicles in comparison to diesel. The primary factors affecting the emissions are the fuel efficiency of natural gas engines relative to diesel and methane leakage across the supply chain.
Article
Engineering, Civil
He-in Cheong, Jose Javier Escribano Macias, Renos Karamanis, Marc Stettler, Arnab Majumdar, Panagiotis Angeloudis
Summary: A single model is crucial for understanding the dynamics of an autonomous ridesharing transport mode from various stakeholders' perspectives, allowing policymakers and companies to develop user-centered strategies. This model should be based on real data and applied to real cities, especially those with complex public transportation systems.
TRANSPORTATION RESEARCH RECORD
(2023)
Article
Biology
Wen-Te Liu, Huei-Tyng Huang, Hsin-Yi Hung, Shang-Yang Lin, Wen-Hua Hsu, Fang-Yu Lee, Yi-Chun Kuan, Yin-Tzu Lin, Chia-Rung Hsu, Marc Stettler, Chien-Ming Yang, Jieni Wang, Ping-Jung Duh, Kang-Yun Lee, Dean Wu, Hsin-Chien Lee, Jiunn-Horng Kang, Szu-Szu Lee, Hsiu-Jui Wong, Cheng-Yu Tsai, Arnab Majumdar
Summary: This study showed that continuous positive airway pressure (CPAP) treatment can alleviate symptoms of obstructive sleep apnea (OSA) and reduce the risk of neurodegenerative diseases. Both sleep quality scores and neurochemical biomarker levels significantly improved in patients treated with CPAP.
Article
Environmental Sciences
Cheng-Yu Tsai, Chien-Ling Su, Yuan-Hung Wang, Sheng-Ming Wu, Wen-Te Liu, Wen-Hua Hsu, Arnab Majumdar, Marc Stettler, Kuan-Yuan Chen, Ya-Ting Lee, Chaur-Jong Hu, Kang-Yun Lee, Ben-Jei Tsuang, Chien-Hua Tseng
Summary: Long-term exposure to air pollution can lead to cardiovascular disease, metabolic syndrome, and chronic respiratory disease. A study in Taiwan evaluated the impact of air pollution exposure at different life stages and found that exposure after 40 years of age may increase the risk of metabolic syndrome, hypertension, diabetes, and cardiovascular disease. Models considering lifetime exposure showed higher precision, accuracy, and F1 scores than models incorporating only late-stage exposures.
ENVIRONMENTAL RESEARCH
(2023)
Article
Engineering, Environmental
Liang Ma, Daniel J. Graham, Marc E. J. Stettler
Summary: By analyzing activity changes during the first UK national lockdown, we found that road traffic significantly decreased and had varying impacts on air pollution in London. The changes in air pollution were correlated with spatial features, residents' income, and access to public transport services. Furthermore, existing inequalities in air pollution exposure were exacerbated during the lockdown.
ENVIRONMENTAL SCIENCE & TECHNOLOGY
(2023)
Article
Environmental Sciences
H. Woodward, A. Schroeder, A. de Nazelle, C. C. Pain, M. E. J. Stettler, H. ApSimon, A. Robins, P. F. Linden
Summary: The spatio-temporal variability of exposure to harmful pollutants in roadside areas is often neglected in assessments of pedestrian and cyclist exposures. This study aims to fully describe this variability and evaluate the benefits of high spatio-temporal resolution over high spatial resolution only. The study also compares high resolution vehicle emissions modeling to using a constant volume source. The findings highlight the impact of peak exposures and emphasize the importance of considering high resolution temporal air pollution variability for accurate characterization of pedestrian and cyclist exposures.
SCIENCE OF THE TOTAL ENVIRONMENT
(2023)
Article
Environmental Studies
Daniel Mehlig, Iain Staffell, Marc Stettler, Helen Apsimon
Summary: The rapid adoption of new vehicle technologies will have a significant impact on the environment in terms of emissions. Specifically, the emissions from power plants supplying electric vehicles (EVs) and non-exhaust PM2.5 emissions from vehicles may undermine the benefits of EVs. A fleet turnover model was developed to assess the effects of different vehicle technologies, the rate of technological change, and changing transport demand on CO2eq and air pollutant emissions. The findings suggest that by 2050, the transition to EVs can reduce annual CO2eq emissions by 98% and cumulative CO2eq emissions by over 50%. Accelerating or delaying the uptake of EVs by 5 years only changes these results by 1% and 17% respectively. Furthermore, EVs can also significantly reduce annual NOx emissions by 97%, but their impact on PM2.5 is limited. Overall, reducing vehicle kilometers has the potential to reduce non-exhaust PM2.5 emissions by 20% in the long term.
TRANSPORTATION RESEARCH PART D-TRANSPORT AND ENVIRONMENT
(2023)
Article
Green & Sustainable Science & Technology
Yingji Xia, Chenlei Liao, Xiqun (Michael) Chen, Zheng Zhu, Xiaorui Chen, Lixing Wang, Rui Jiang, Marc E. J. Stettler, Panagiotis Angeloudis, Ziyou Gao
Summary: This paper examines how individual drivers can contribute to reducing global vehicle emissions by modifying their driving behavior. The study proposes a Standardized Driver Aggressiveness Index to estimate changes in driving behavior based on real-world trajectory data. The research predicts the additional vehicle emissions induced by different types of car-following behavior and finds that cumulative emissions that could be prevented by 2050 amount to 400.5 million tons of CO2. The findings emphasize the importance of considering behavioral changes in mitigating transport emissions and the need for interventions to encourage sustainable driving behavior.
NATURE SUSTAINABILITY
(2023)
Article
Health Care Sciences & Services
Cheng-Yu Tsai, Wen-Te Liu, Wen-Hua Hsu, Arnab Majumdar, Marc Stettler, Kang-Yun Lee, Wun-Hao Cheng, Dean Wu, Hsin-Chien Lee, Yi-Chun Kuan, Cheng-Jung Wu, Yi-Chih Lin, Shu-Chuan Ho
Summary: This study aimed to establish machine learning models to screen for the risk of moderate-to-severe and severe obstructive sleep apnea (OSA) based on easily acquired features. Data from 3529 patients in Taiwan were collected, and six common supervised machine learning techniques were utilized. The random forest (RF) produced the highest accuracy (>70%) in screening for both OSA severities.
Article
Health Care Sciences & Services
Chih-Fan Kuo, Cheng-Yu Tsai, Wun-Hao Cheng, Wen-Hua Hs, Arnab Majumdar, Marc Stettler, Kang-Yun Lee, Yi-Chun Kuan, Po-Hao Feng, Chien-Hua Tseng, Kuan-Yuan Chen, Jiunn-Horng Kang, Hsin-Chien Lee, Cheng-Jung Wu, Wen-Te Liu
Summary: This study used easy-to-measure parameters to predict sleep arousal and provided a feasible model for screening sleep arousal occurrence.
Article
Engineering, Civil
Qiming Ye, Yuxiang Feng, Jose Javier Escribano Macias, Marc Stettler, Panagiotis Angeloudis
Summary: The deployment of Autonomous Vehicles (AVs) presents challenges and opportunities for the design and management of urban road infrastructure. This study explores the evolution of road Right-Of-Way (ROW) using Reinforcement Learning (RL) methods, implementing both centralized and distributed learning paradigms for dynamic control of road networks. Experimental results show that these algorithms can improve traffic flow efficiency and allocate more space for pedestrians.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)