Sort by
Machine Learning Strategies for Forecasting Mannosylerythritol Lipid Production in Fermentation Processes
Carolina A. Vares,
Sofia P. Agostinho,
Ana L.N. Fred,
Nuno T. Faria,
Carlos A. V. Rodrigues
Fermentations are complex and often unpredictable processes. However, fermentation-based bioprocesses generate large volumes of data that are currently underexplored. These data can be used to develop data-driven models, such as machine learning (ML), to improve process predictability. Among various fermentation products, biosurfactants have emerged as promising candidates for several industrial applications. Nevertheless, biosurfactant large-scale production is not yet cost-effective. This study aims to develop forecasting methods for the concentration of mannosylerythritol lipids (MELs), a type of biosurfactant, produced in Moesziomyces spp. cultivation. Three ML models, Neural Networks (NN), Support Vector Machines (SVM), and Random Forests (RF), were used. NN provided predictions with a mean squared error (MSE) of 0.69 for day 4 and 1.63 for day 7, and a mean absolute error (MAE) of 0.58 g/L and 1.1 g/L, respectively. These results indicate that the model’s predictions are sufficiently accurate for practical use, with the MAE showing only minor deviations from the actual concentrations. Both results are promising, as they demonstrate the possibility of obtaining reliable predictions of MELs production for days 4 and 7 of fermentation. This, in turn, could help reduce process-related costs, enhancing its economic viability.
Fermentations are complex and often unpredictable processes. However, fermentation-based bioprocesses generate large volumes of data that are currently underexplored. These data can be used to develop data-driven models, such as machine learning (ML), to improve process predictability. Among various fermentation products, biosurfactants have emerged as promising candidates for several industrial applications. Nevertheless, biosurfactant large-scale production is not yet cost-effective. This study aims to develop forecasting methods for the concentration of mannosylerythritol lipids (MELs), a type of biosurfactant, produced in Moesziomyces spp. cultivation. Three ML models, Neural Networks (NN), Support Vector Machines (SVM), and Random Forests (RF), were used. NN provided predictions with a mean squared error (MSE) of 0.69 for day 4 and 1.63 for day 7, and a mean absolute error (MAE) of 0.58 g/L and 1.1 g/L, respectively. These results indicate that the model’s predictions are sufficiently accurate for practical use, with the MAE showing only minor deviations from the actual concentrations. Both results are promising, as they demonstrate the possibility of obtaining reliable predictions of MELs production for days 4 and 7 of fermentation. This, in turn, could help reduce process-related costs, enhancing its economic viability.
Posted: 14 February 2025
Revolutionizing Medical Image Segmentation: A Deep Dive into Challenges and Future of Federated Learning
Saba Salahuddin,
Khan Bahadar Khan,
Abdul Qayyum,
Iftikhar Ahmed,
Kashif Saleem,
Sadia Saeed
The possibility of medical image segmentation within the domain of a federated learning, Federated Learning (FL) may transform the situation and help solve the critical challenges that exist in common centralized machine learning models. While effective, traditional models are limited by issues like the need of huge surveys, high costs in data assignment, high privacy concerns over sensible wellbeing data. Since improvements in the medical imaging field continue, the adoption of FL is a strategic response to such limitations and can be introduced as a collaborative privacy preserving framework for model training. This was a systematic exploration of the literature from 2017 to 2024 where the Google Scholar literature has been explored for studies indexed with the keywords 'federated learning,' 'medical image segmentation,' and 'privacy preservation.' Specifically, this review did not consider studies that did not directly discuss FL concepts. Twenty-one publications were carefully selected from out of thousands of publications because they are relevant and contribute to the area of treatment. Specifically, seven studies directly approached the extent of medical image segmentation using FL and address the technological and the practical challenges. The remaining fourteen studies were foundational in that they further elaborated on the architectural and procedural elements of FL frameworks that are essential for collaborative and secure medical image analysis. A review of the selected studies is presented in detail in the review in terms of the effectiveness of FL in improving medical image segmentation while protecting patient privacy. It makes a powerful evaluation of the strengths and weakness of present FL model, the versatility of data sets, the diversity of the imaging modalities addressed, and scalability of these models across various clinical conditions. Such synthesis of this literature underscores the fact that FL can revolutionize medical diagnostics with opportunity to produce more robust, scalable, and privacy friendly models.
The possibility of medical image segmentation within the domain of a federated learning, Federated Learning (FL) may transform the situation and help solve the critical challenges that exist in common centralized machine learning models. While effective, traditional models are limited by issues like the need of huge surveys, high costs in data assignment, high privacy concerns over sensible wellbeing data. Since improvements in the medical imaging field continue, the adoption of FL is a strategic response to such limitations and can be introduced as a collaborative privacy preserving framework for model training. This was a systematic exploration of the literature from 2017 to 2024 where the Google Scholar literature has been explored for studies indexed with the keywords 'federated learning,' 'medical image segmentation,' and 'privacy preservation.' Specifically, this review did not consider studies that did not directly discuss FL concepts. Twenty-one publications were carefully selected from out of thousands of publications because they are relevant and contribute to the area of treatment. Specifically, seven studies directly approached the extent of medical image segmentation using FL and address the technological and the practical challenges. The remaining fourteen studies were foundational in that they further elaborated on the architectural and procedural elements of FL frameworks that are essential for collaborative and secure medical image analysis. A review of the selected studies is presented in detail in the review in terms of the effectiveness of FL in improving medical image segmentation while protecting patient privacy. It makes a powerful evaluation of the strengths and weakness of present FL model, the versatility of data sets, the diversity of the imaging modalities addressed, and scalability of these models across various clinical conditions. Such synthesis of this literature underscores the fact that FL can revolutionize medical diagnostics with opportunity to produce more robust, scalable, and privacy friendly models.
Posted: 14 February 2025
Mobile Phone Studies Find No Short-Term Health Problems
Georgios Giannakopoulos,
Khushbu Mehboob Shaikh,
Maria Antonnette Perez
Posted: 13 February 2025
Phased Array Antennas: Advancements and Applications
Georgios Giannakopoulos,
Khushbu Mehboob Shaikh
Phased array antennas provide the ability to electronically steer a beam, eliminating the need for mechanical adjustments [1]. While traditionally used in military applications, there is growing interest in their adoption across various fields [1,2]. Conformal antennas, a type of phased array, are designed for installation on curved or non-flat surfaces, enabling focused radio wave radiation [1,2]. These antennas can be integrated into various applications, including aerospace, wearable technology, vehicles, and modern mobile devices [2], while also reducing traditional antenna height to support the integration and coexistence of multiple radio technologies within a compact area [1,2]. Planar arrays, composed of elements with phase shifters in a matrix, are compact and cost-effective due to mass production via printed circuit technology [1–3]. These antennas, when mounted on rigid surfaces, exhibit robustness, provide beam deflection in two planes, and offer high gain with rapid beam-switching capabilities [1,3]. However, planar antennas can experience interference between feed lines and elements, often supporting narrow bandwidths and exhibiting relatively low radiation efficiency [1,3]. Conformal antennas, which are easily mounted on curved surfaces, are particularly suited for wearable applications, spacesuits, and aerospace designs [1,2,4]. By minimizing connection length, they bring electronics closer to the antenna elements, reducing signal loss while enhancing transmission power and receiver sensitivity, especially at higher frequencies [4]. Research into 3Dprinted conformal antennas has emerged as a significant field of study [1,5]. This paper presents the mathematical analysis of both planar and conformal antennas, covering key parameters such as gain, bandwidth, radiation efficiency, and mutual coupling for planar arrays, as well as the width and length calculations for rectangular microstrip patch antennas used in conformal designs [2,6–8]. Furthermore, the role of additive manufacturing in antenna development is highlighted, emphasizing its ability to produce antennas with complex geometries thereby revolutionizing conformal antenna design [1,9].
Phased array antennas provide the ability to electronically steer a beam, eliminating the need for mechanical adjustments [1]. While traditionally used in military applications, there is growing interest in their adoption across various fields [1,2]. Conformal antennas, a type of phased array, are designed for installation on curved or non-flat surfaces, enabling focused radio wave radiation [1,2]. These antennas can be integrated into various applications, including aerospace, wearable technology, vehicles, and modern mobile devices [2], while also reducing traditional antenna height to support the integration and coexistence of multiple radio technologies within a compact area [1,2]. Planar arrays, composed of elements with phase shifters in a matrix, are compact and cost-effective due to mass production via printed circuit technology [1–3]. These antennas, when mounted on rigid surfaces, exhibit robustness, provide beam deflection in two planes, and offer high gain with rapid beam-switching capabilities [1,3]. However, planar antennas can experience interference between feed lines and elements, often supporting narrow bandwidths and exhibiting relatively low radiation efficiency [1,3]. Conformal antennas, which are easily mounted on curved surfaces, are particularly suited for wearable applications, spacesuits, and aerospace designs [1,2,4]. By minimizing connection length, they bring electronics closer to the antenna elements, reducing signal loss while enhancing transmission power and receiver sensitivity, especially at higher frequencies [4]. Research into 3Dprinted conformal antennas has emerged as a significant field of study [1,5]. This paper presents the mathematical analysis of both planar and conformal antennas, covering key parameters such as gain, bandwidth, radiation efficiency, and mutual coupling for planar arrays, as well as the width and length calculations for rectangular microstrip patch antennas used in conformal designs [2,6–8]. Furthermore, the role of additive manufacturing in antenna development is highlighted, emphasizing its ability to produce antennas with complex geometries thereby revolutionizing conformal antenna design [1,9].
Posted: 13 February 2025
Impact of Exhaust Manifold Design on Internal Combustion Engine Performance
Darioush Jamshidi,
Daniyal Poureyvaz Borazjani,
Seyed Ehsan Hosseini,
Sajad Davari
Posted: 13 February 2025
The Design of a Vision-assisted Dynamic Antenna Positioning RFID-based Inventory Robot Utilizing a 3DOF Manipulator
Abdussalam A. Alajami,
Rafael Pous
Posted: 13 February 2025
Peculiarities of Soil Tillage in Southeastern Kazakhstan
Askar Rzaliyev,
Valeriya Goloborodko,
Serik Bekbosynov,
Olzhas Seipataliyev,
Dauren Kosherbay
Posted: 13 February 2025
A Survey of Route Bus Speed Change Pattern for Clarifying Electrification Benefits
Yiyuan Fang,
Wei-hsiang Yang,
Yushi Kamiya
Posted: 13 February 2025
Use of Alternative Materials in Sustainable Geotechnics: State of World Knowledge and Some Examples from Poland
Małgorzata Jastrzębska
Posted: 13 February 2025
Emotional Modulation of Decision Processes in Coal Mine Emergencies: ERP Insights into Miners’ Crisis Response
Yifan Zhao,
Shuicheng Tian,
Junrui Mao,
Guangtong Shao
Posted: 13 February 2025
The Development of a Nitinol Angular Actuator
Oana Vasilica Grosu,
Laurențiu Dan Milici,
Mihaela Paval
Posted: 13 February 2025
Anomaly Monitoring Model of Industrial Processes Based on Graph Similarity and Applications
Guoqing Du,
Mingyi Yang,
Zhigang Xu,
Junyi Wang,
Cheng Xie,
Yuan Lu,
Pengfei Yin
Posted: 13 February 2025
Framework for the Multi-Objective Design Optimization of Aerocapture Missions
Segundo Urraza Atue,
Paul Bruce
Posted: 13 February 2025
Numerical Method to Obtain the Resistance of Reinforced Concrete and Composite Steel Concrete Sections Subjected to Axial Forces and Biaxial Bending Based on Simplex (Linear Programming)
Antonio Aguero,
Ivan Baláž,
Yvona Kolekova
Posted: 13 February 2025
Research Hardware Publications Should Help Build Communities, Careers and Better Hardware
Julien Colomb,
Moritz Maxeiner,
Robert Mies
Based on the interviews of fifteen contributors in research hardware projects, we shed lights onto the motivations of research hardware engineers to use a hardware publication platform, and derived some high-level features that such a platform should have. Our analysis suggests that the main objectives of the authors are to find their readership and grow an inclusive community of contributors, producers and users around their project. Inclusivity requires the recognition of different types of contributions and a system free of financial or language barriers for authors and readers. They also appear interested in getting feedback on their work, in order to make the hardware better and learn during that process. The creation of a research output that is recognized by the academic system is also important both for their career and for developing their community. In addition, they express wishes for a publication system integrated in their hardware documentation workflow , as well as a system which would be pleasant or even fun to use. Importantly, a research hardware publication ecosystem should link archived and living versions of the hardware project and consider the project as a whole, providing documentation on both the hardware product and the development process.
Based on the interviews of fifteen contributors in research hardware projects, we shed lights onto the motivations of research hardware engineers to use a hardware publication platform, and derived some high-level features that such a platform should have. Our analysis suggests that the main objectives of the authors are to find their readership and grow an inclusive community of contributors, producers and users around their project. Inclusivity requires the recognition of different types of contributions and a system free of financial or language barriers for authors and readers. They also appear interested in getting feedback on their work, in order to make the hardware better and learn during that process. The creation of a research output that is recognized by the academic system is also important both for their career and for developing their community. In addition, they express wishes for a publication system integrated in their hardware documentation workflow , as well as a system which would be pleasant or even fun to use. Importantly, a research hardware publication ecosystem should link archived and living versions of the hardware project and consider the project as a whole, providing documentation on both the hardware product and the development process.
Posted: 13 February 2025
Design and Characterization of the Modified Purdue Subcritical Pile for Nuclear Research Applications
Matthew Niichel,
Riley Madden,
Hannah Pike,
Nafees Bin Kabir,
True Miller,
Brian Jowers,
Stylianos Chatzidakis
First demonstrated in 1942, subcritical and zero-power critical assemblies, also known as piles, are a fundamental tool for research and education at universities. Traditionally, their role has been primarily instructional and for measuring fundamental properties of neutron diffusion and transport. However, these assemblies could hold potential for modern applications and nuclear research. The Purdue University subcritical pile previously lacked a substantial testing volume, limiting its utility to simple neutron activation experiments for the purpose of undergraduate education. Following the design and addition of a mechanical and electrical testbed, this paper aims to provide an overview of the testbed design and characterize its neutron and gamma flux of the rearranged Purdue subcritical pile, justifying its use as a modern scientific instrument. The newly installed 1.5*10^5 cubic-centimeter volume testbed enables a systematic investigation of neutron and gamma effects on materials and the generation of a comprehensive dataset with the potential for machine learning applications. The neutron flux throughout the pile is calculated using gold-197 and indium-115 foil activation alongside cadmium-covered foils for two-group neutron energy classification. The neutron flux measurements are then used to benchmark a detailed geometrically and materialistic accurate Monte-Carlo model using OpenMC. The experimental measurements reveal the testbed has a neutron environment with a total neutron flux approaching 8.5*10^3 n/cm^2*s and a thermal flux of 5.8*10^3 n/cm^2*s, following the expected diffusion theory behavior. This work establishes the modified Purdue subcritical pile can provide significant neutron and gamma fluxes and a uniquely large volume to enable radiation testing of integral electronic components and as a versatile research instrument with the potential to support microelectronics testing, limited isotope production, and non-destructive imaging while generating valuable training datasets for machine learning algorithms in nuclear applications.
[M1]Reference citation is not allowed. Please revise.
First demonstrated in 1942, subcritical and zero-power critical assemblies, also known as piles, are a fundamental tool for research and education at universities. Traditionally, their role has been primarily instructional and for measuring fundamental properties of neutron diffusion and transport. However, these assemblies could hold potential for modern applications and nuclear research. The Purdue University subcritical pile previously lacked a substantial testing volume, limiting its utility to simple neutron activation experiments for the purpose of undergraduate education. Following the design and addition of a mechanical and electrical testbed, this paper aims to provide an overview of the testbed design and characterize its neutron and gamma flux of the rearranged Purdue subcritical pile, justifying its use as a modern scientific instrument. The newly installed 1.5*10^5 cubic-centimeter volume testbed enables a systematic investigation of neutron and gamma effects on materials and the generation of a comprehensive dataset with the potential for machine learning applications. The neutron flux throughout the pile is calculated using gold-197 and indium-115 foil activation alongside cadmium-covered foils for two-group neutron energy classification. The neutron flux measurements are then used to benchmark a detailed geometrically and materialistic accurate Monte-Carlo model using OpenMC. The experimental measurements reveal the testbed has a neutron environment with a total neutron flux approaching 8.5*10^3 n/cm^2*s and a thermal flux of 5.8*10^3 n/cm^2*s, following the expected diffusion theory behavior. This work establishes the modified Purdue subcritical pile can provide significant neutron and gamma fluxes and a uniquely large volume to enable radiation testing of integral electronic components and as a versatile research instrument with the potential to support microelectronics testing, limited isotope production, and non-destructive imaging while generating valuable training datasets for machine learning algorithms in nuclear applications.
[M1]Reference citation is not allowed. Please revise.
Posted: 13 February 2025
Class 4 Sections Computing Section Properties and Stresses Due to the Bimoment
Antonio Aguero,
Ivan Baláž,
Yvona kolekova
Posted: 13 February 2025
Smart Building Technologies for Fire Rescue: A QR Code Enabled Notification System
Tzu-Wen Kuo,
Ching-Yuan Lin
This study aimed to shorten firefighter search times during indoor fires, allowing more people to be rescued, by enhancing disaster prevention capabilities using building technologies. In indoor fires, fatalities are often caused by the failure of firefighters to rescue individuals in a timely manner. The question of how to effectively increase the probability of survival while waiting for rescue behind closed doors warrants in-depth research and analysis. Therefore, to ensure that people live in safe environments, there is an urgent need to develop a building door panel material with an emergency call function to prevent such incidents from occurring. Utilizing the PRISMA method, we conducted a comprehensive review of the existing literature to identify the key issues and limitations associated with the current search-and-rescue techniques. Subsequently, the identified primary factors were analyzed using the TRIZ method to determine the key factors that influence the success of rescuing trapped individuals, and a notification system was designed to address this issue. Based on the premise that it is advisable to wait for rescue during a fire, we utilized a smartphone to scan a QR code and transmit the exact location information to the fire department. Through extensive participation and feedback from firefighters, we developed a rescue notification door panel and obtained a patent for it. This system can significantly reduce the time required for search-and-rescue operations in fire incidents. The experimental results show a reduction of one-third in search times.
This study aimed to shorten firefighter search times during indoor fires, allowing more people to be rescued, by enhancing disaster prevention capabilities using building technologies. In indoor fires, fatalities are often caused by the failure of firefighters to rescue individuals in a timely manner. The question of how to effectively increase the probability of survival while waiting for rescue behind closed doors warrants in-depth research and analysis. Therefore, to ensure that people live in safe environments, there is an urgent need to develop a building door panel material with an emergency call function to prevent such incidents from occurring. Utilizing the PRISMA method, we conducted a comprehensive review of the existing literature to identify the key issues and limitations associated with the current search-and-rescue techniques. Subsequently, the identified primary factors were analyzed using the TRIZ method to determine the key factors that influence the success of rescuing trapped individuals, and a notification system was designed to address this issue. Based on the premise that it is advisable to wait for rescue during a fire, we utilized a smartphone to scan a QR code and transmit the exact location information to the fire department. Through extensive participation and feedback from firefighters, we developed a rescue notification door panel and obtained a patent for it. This system can significantly reduce the time required for search-and-rescue operations in fire incidents. The experimental results show a reduction of one-third in search times.
Posted: 13 February 2025
The Frequency Response Characteristics of Ge-on-Si Photodetectors Under High Incident Power
Jin Jiang,
Hongmin Chen,
Fenghe Yang,
Chunlai Li,
Jin He,
Xiumei Wang,
Jishi Cui
Posted: 13 February 2025
A Machine Learning Classification Approach to Geotechnical Characterisation Using Measure-While-Drilling Data
Daniel Goldstein,
Chris Aldrich,
Quanxi Shao,
Louisa O'Connor
Posted: 13 February 2025
of 608