20505 research outputs found
Sort by
Π-ORFit: One-Pass Learning with Bregman Projection
Get PDFThis paper delves into the problem of one-pass learning, where the objective is to train a model on each datapoint in a stream while maintaining performance on past data without retraining on them. An existing approach to this problem in the context of overparameterized (underdetermined) models is Orthogonal Recursive Fitting (ORFit), which fits every new data point while maintaining predictions on previous datapoints by ensuring that parameter updates are orthogonal to the directions that are critical for past data (i.e., the direction of gradient of the model output with respect to the parameters, for those data). For overparameterized linear models, when initialized at zero, ORFit obtains the parameter vector that perfectly fits the data and has the minimum ℓ2 -norm, among the infinitely many perfectly fitting parameter vectors. To generalize this and gain control over the selection of desired parameters, in this paper, we introduce Projected Orthogonal Recursive Fitting (Π-ORFit). We begin by characterizing all parameters that can precisely fit data in general vector-output linear models, employing a formalism based on nullspace projector matrices. This framework yields an alternative derivation of ORFit. Building on this, we further extend ORFit to learn a desired parameter by incorporating a Bregman projection into the update rule. Importantly, we show that the resulting parameter minimizes the potential function that defines the Bregman projection at each update step, enabling the selection of a desired parameter among the (infinitely many) candidates consistent with the data. We provide numerical experiments that validate our analytical findings and underscore the practical significance of this generalized approach.2024 American Control Conference (ACC
Demonstration of anaerobic membrane bioreactors for resource recovery in wastewater treatment applications.
Get PDFJefferson, Bruce - Associate SupervisorPilot-scale studies on anaerobic membrane bioreactors (AnMBRs) for municipal
wastewater treatment at low temperature (<20°C) showed promising results,
however, their application at larger scale is still relatively unknown. This study
investigated the scalability of an AnMBR, comprising an upflow anaerobic sludge
blanket (UASB) reactor and an external ultrafiltration membrane tank, operating
AnMBRs both at pilot- and demonstration-scale and identifying how sludge
physical and microbiological properties, membrane design and variations in
influent temperature, chemical oxygen demand (COD) and sulphate (SO₄)
influence the successful scale-up of the technology. At pilot-scale, the source and
adaptation of the inoculum and the orientation and arrangement of the membrane
fibres did not affect the performance of the reactors. However, the use of
horizontal hollow fibres led to lower gas sparging energy consumption compared
to a vertical module. The membrane improved removal efficiencies (from 49-57%
to 88-92% COD removal), solids hydrolysed (from 0.82-0.86 g/(L∙d) to 1.57-1.87
g/(L∙d)) and methane production (from 2.3-2.7 L/d to 5.3-5.7 L/d).
Methanogenesis percentages were linearly correlated to hydrolysis, which in turn
was affected by temperature and inversely correlated to the Sauter mean
diameter of the sludge particles. Higher substrate affinities were found at the
operational temperature of the reactors (15-20°C), while hydrolytic enzyme
activities in UASB reactors and AnMBRs were higher at 37°C. Methane was
mainly dissolved in the effluent (70-90%), implying the need for a recovery
process to improve the net energy balance. At demonstration-scale, low
COD:SO4 ratio caused competition between sulphate-reducing bacteria and
methanogens, leading to a decrease in methane yield. This study proved that
AnMBRs are a suitable technology to treat municipal wastewater, however site-
specific control strategies to manage fouling and sulphate and appropriate post-
treatments are necessary to ensure the successful application of the process at
full-scale in temperate climates and the recovery of useful resources from
wastewater.PhD in Water, including Desig
Recognition of Brazilian vertical traffic signs and lights from a car using Single Shot Multi box Detector
Get PDFThis work presents an automated system for recognizing Brazilian vertical traffic signs and lights using artificial intelligence. The main objective of the system is to contribute to road safety by alerting drivers to potential risks such as speeding, alcohol consumption, and cell phone use, which could lead to severe accidents. The system’s core contribution lies in its ability to accurately recognize various traffic signs and lights, providing crucial warnings to drivers. To achieve this, the system utilizes a light version of the single shot multi box detector as its detection algorithm and experiments with three Mobilenet versions as base networks. The optimal Mobilenet version is selected based on a mean average precision higher than 80%, which guarantees reliable detection results. The dataset used for training and evaluation comprises images extracted from YouTube traffic videos, each annotated to create the necessary labels for training. Through this extensive experimentation, the system demonstrates its efficacy in achieving accurate and efficient detection. The results of the experiments are compared with other existing approaches and our work significantly advances the field by providing a tailored dataset, an optimized model, and also valuable insights into traffic sign and light recognition, collectively contributing to the improvement of road safety.Journal of the Brazilian Computer Societ
Application of machine learning in hydrogen production via the process of sorbent enhanced steam methane reforming.
Get PDFManovic, Vasilije - Associate Supervisor
Wagland, Stuart - Associate SupervisorThis thesis is focused on the exploration of the use of machine learning and
computational methods for modelling process conditions and for materials
screening within the process of sorbent enhanced steam methane reforming (SE-
SMR) for carbon-abated hydrogen production. Hydrogen is a clean, abundant
and versatile energy carrier that can be used for a wide range of applications.
However, the production of hydrogen is still largely dependent on fossil fuels,
which presents a significant challenge for achieving a truly sustainable energy
system.
The purpose of this study is to address this challenge by exploring novel
approaches to hydrogen production, namely using machine learning,
thermodynamic simulations, theoretical modelling, and the proposal of new
methodologies and materials for low-carbon hydrogen production.
Three main areas of work were conducted within this thesis, which include 1) two
surrogate models have been developed and used to predict and estimate
variables that would otherwise be difficult direct measured.; 2) applying machine
learning, namely quantitative structure–property relationship analysis (QSPR)
has been employed in the exploration of combined sorbent catalyst material
(CSCM) for SE-SMR; and 3) applying machine learning to screen suitable metal
organic frameworks (MOFs) for the storage of the produced blue hydrogen.
Firstly, a surrogate model, was developed which was done by firstly simulating
the model in Aspen Plus, applying a sensitivity analysis to gather a large dataset,
then applying two multiple linear regression model, to observe the accuracy of
predicting the gas concentration outputs. Two models were successfully
developed with both models were accurate with high R² values, all above 98%.
Secondly, the novel approach of QSPR with inductive transfer learning and
datamining, was applied to develop two large databases of sorbent and catalyst
properties, respectively. Then the developed machine learning models from
these databases were applied, to predict the optimal conditions and precursor
materials for the highest performing CSCM, in terms of last cycle capacity and
methane conversion. Lastly, a similar approach was applied for the screening of
MOFs for the storage of hydrogen by using multiple linear regression, simple
geometric descriptors, and patterns in data to identify a better performing MOF
than the currently reported experimental MOFs in literature.PhD in Energy and Powe
Detection of sugar syrup adulteration in UK honey using DNA barcoding
Get PDFHoney is a valuable and nutritious food product, but it is at risk to fraudulent practices such as the addition of cheaper syrups including corn, rice, and sugar beet syrup. Honey authentication is of the utmost importance, but current methods are faced with challenges due to the large variations in natural honey composition (influenced by climate, seasons and bee foraging), or the incapability to detect certain types of plant syrups to confirm the adulterant used. Molecular methods such as DNA barcoding have shown great promise in identifying plant DNA sources in honey and could be applied to detect plant-based sugars used as adulterants. In this work DNA barcoding was successfully used to detect corn and rice syrup adulteration in spiked UK honey with novel DNA markers. Different levels of adulteration were simulated (1 – 30%) with a range of different syrup and honey types, where adulterated honey was clearly separated from natural honey even at 1% adulteration level. Moreover, the test was successful for multiple syrup types and effective on honeys with different compositions. These results demonstrated that DNA barcoding could be used as a sensitive and robust method to detect common sugar adulterants and confirm syrup species origin in honey, which can be applied alongside current screening methods to improve existing honey authentication tests.Science and Technology Facilities Council (STFC)Biotechnology and Biological Sciences Research Council (BBSRC)This research was funded by the UKRI STFC Food Network+ [Grant No: ST/T002921/1], the Food Standards Agency (FSA) [Project No: FS900185] and UKRI BBSRC FoodBioSystems Doctoral Training Partnership (DTP) [Grant No: BB/T008776/1].Food Contro
The influence of different abiotic conditions on the concentrations of free and conjugated deoxynivalenol and zearalenone in stored wheat
Get PDFEnvironmental factors influence fungal growth and mycotoxin production in stored grains. However, the concentrations of free mycotoxins and their conjugates and how they are impacted by different interacting environment conditions have not been previously examined. The objectives of this study were to examine the impact of storage conditions (0.93–0.98 aw) and temperature (20–25 °C) on (a) the concentrations of deoxynivalenol and zearalenone and their respective glucosides/conjugates and (b) the concentrations of emerging mycotoxins in both naturally contaminated and irradiated wheat grains inoculated with Fusarium graminearum. Contaminated samples were analysed for multiple mycotoxins using Liquid Chromatography Tandem Mass Spectrometry (LC–MS/MS). Method validation was performed according to the acceptable performance criteria set and updated by the European Commission regulations No. 2021/808/EC. As an important conjugate of deoxynivalenol, the concentrations of deoxynivalenol-3-glucoside were significantly different from its precursor deoxynivalenol at 0.93 aw (22% moisture content- MC) at 25 °C in the naturally contaminated wheat with a ratio proportion of 56:44% respectively. The high concentrations of deoxynivalenol-3-glucoside could be influenced by the wheat’s variety and/or harvested season/fungal strain type/location. Zeralenone-14-sulfate concentrations were surprisingly three times higher than Zearalenone in the naturally contaminated wheat at 0.98 aw (26% MC) at both temperatures. Emerging mycotoxins such as moniliformin increased with temperature rise with the highest concentrations at 0.95 aw and 25 °C. These findings highlight the influence and importance of storage aw x temperature conditions on the relative presence of free vs conjugated mycotoxins which can have implications for food safety.Biotechnology and Biological Sciences Research Council (BBSRC)This research was funded by the United Kingdom Research and Innovation (UKRI), the Biotechnology and Biological Sciences Research Council (BBSRC), and the FoodBioSystems Doctoral Training Partnership (FBSDTP), grant number: BB/T008776/1.Mycotoxin Researc
Nanomaterial integration in micro LED technology: enhancing efficiency and applications
Get PDFThe micro-light emitting diode (µLED) technology is poised to revolutionise display applications through the introduction of nanomaterials and Group III-nitride nanostructures. This review charts state-of-the-art in this important area of micro-LEDs by highlighting their key roles, progress and concerns. The review encompasses details from various types of nanomaterials to the complexity of gallium nitride (GaN) and III nitride nanostructures. The necessity to integrate nanomaterials with III-nitride structures to create effective displays that could disrupt industries was emphasised in this review. Commercialisation challenges and the economic enhancement of micro-LED integration into display applications using monolithic integrated devices have also been discussed. Furthermore, different approaches in micro-LED development are discussed from top-down and bottom-up approaches. The last part of the review focuses on nanomaterials employed in the production of micro-LED displays. It also highlights the combination of III-V LEDs with silicon LCDs and perovskite-based micro-LED displays. There is evidence that efficiency and performance have improved significantly since the inception of the use of nanomaterials in manufacturing these.Raghvendra Kumar Mishra would like to acknowledge the financial support provided by the UKRI. The research received funding from the Engineering & Physical Sciences Research Council (EPSRC), UK – Ref. EP/R016828/1 (Self-tuning Fibre-Reinforced Polymer Adaptive Nanocomposite, STRAIN comp) and EP/R513027/1 (Study of Microstructure of Dielectric Polymer Nanocomposites subjected to Electromagnetic Field for Development of Self-toughening Lightweight Composites). SG would like to acknowledge the financial support provided by the UKRI via Grants No. EP/S036180/1 and EP/T024607/1, Hubert Curien Partnership Programme from the British Counsil and the International exchange Cost Share award by the Royal Society (IEC\NSFC\223536).Next Nanotechnolog
Comparative sanitation data from high-frequency phone surveys across 3 countries
Get PDFWith less than half of the worldʼs urban population having safely managed sanitation due to the high cost and difficulty of building sewers and treatment plants, many rely on off-grid options like pit latrines and septic tanks, which are hard to empty and often lead to illegal waste dumping; this research focuses on container-based sanitation (CBS) as an emerging off-grid solution. Off-grid sanitation refers to waste management systems that operate independently of centralized infrastructure and CBS is a service providing toilets that collect human waste in sealable containers, which are regularly emptied and safely disposed of. These data relate to a project investigating CBS in Kenya, Peru, and South Africa, focusing on how different user groups access and utilize sanitation – contrasting CBS with other types. Participants, acting as citizen scientists, collected confidential data through a dedicated smartphone app designed by the authors and external contractors. This project aimed to explore the effective scaling, management, and regulation of off-grid sanitation systems, relevant to academics in urban planning, water and sanitation services, institutional capability, policy and governance, and those addressing inequality and poverty reduction.
The 12-month data collection period offered participants small incentives for weekly engagement, in a micro payment for micro tasks approach. Participants were randomly selected, attended a training workshop, and (where needed) were given a smartphone which they could keep at the end of the project. We conducted weekly smartphone surveys in over 300 households across informal settlements. These surveys aimed to understand human-environment interactions by capturing daily life, wellbeing, income, infrastructural service use, and socioeconomic variables at a weekly resolution, contributing to more informed analyses and decision-making.
The smartphone-based approach offers efficient, cost-effective, and flexible data collection, enabling extensive geographical coverage, broad subject areas, and frequent engagement. The Open Data Kit (ODK) tools were used to support data collection in the resource-constrained environment with limited or intermittent connectivity.This work took place under the ‘Scaling-up Off-grid Sanitation’ project (SOS; ES/T007877/1), funded with support from the United Kingdom's Global Challenge Research Fund, via the Economic and Social Research Council.Data in Brie
Wildfire and smoke early detection for drone applications: a light-weight deep learning approach
Get PDFDrones have become a crucial element in current wildfire and smoke detection applications. Several deep learning architectures have been developed to detect fire and smoke using either colour-based methodologies or semantic segmentation techniques with impressive results. However, the computational demands of these models reduce their usability on memory-restricted devices such as drones. To overcome this memory constraint whilst maintaining the high detection capabilities of deep learning models, this paper proposes two lightweight architectures for fire and smoke detection in forest environments. The approaches use the Deeplabv3+ architecture for image segmentation as baseline. The novelty lies in the incorporation of vision transformers and a lightweight convolutional neural network architecture that heavily reduces the model complexity, whilst maintaining state-of-the-art performance. Two datasets for fire and smoke segmentation, based on the Corsican, FLAME, SMOKE5K, and AI-For-Mankind datasets, are created to cover different real-world scenarios of wildfire to produce models with better detection capabilities. Experiments are conducted to show the benefits of the proposed approach and its relevance in current drone-based wildfire detection applications.Engineering Applications of Artificial Intelligenc
Removal of perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid (PFOS) by coagulation: influence of coagulant and dosing conditions
Get PDFPer- and polyfluoroalkyl substances (PFAS) pose significant risks to the environment and human health. Perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid (PFOS) are two of the most frequently detected PFAS in the environment. In most surface water drinking water treatment works (WTW), coagulation is the first processes exposed to a range of contaminants, including PFAS. While not designed to be a process for removal of micropollutants, it is important to understand the fate of PFAS in coagulation processes, intended or otherwise, to determine whether water treatment sludge can be a significant sink for this group of micropollutants. This work advances understanding of PFAS removal in coagulation processes by comparing the removal of PFOA and PFOS by four metal coagulants (Zr, Zn, Fe, and Al) from real water matrices. The coagulant performance followed the order Al > Fe > Zr > Zn. Al was taken forward for further evaluation, with significant removal of PFAS (>15 % for PFOA and > 30 % for PFOS) being observed when the pH 5 mg Al·L-1. The adsorption of PFOA and PFOS onto flocs through hydrophobic interaction was the primary removal route. The impacts of background matrix on the mechanisms of coagulation for PFAS were explored using five organic compounds. Macromolecular organic compounds contributed to an increase in removal due to the sorption of PFAS and subsequent removal of the organic-PFAS aggregate during coagulation. Low molecular weight organic matter inhibited the removal of PFAS due to the ineffective removal of these compounds during coagulation.China Scholarship CouncilThe authors would like to express their gratitude for the financial support of the work from the China Scholarship Council.Separation and Purification Technolog