110830 research outputs found
Sort by
Advancing Efficiency in PVT Solar Technology by Leveraging Artificial Intelligence in Intelligent Thermal Management
Get PDFPhotovoltaic-Thermal (PVT) systems have a strong potential to improve solar technology in energy generation and conversion. The performance of PVT systems is, however, critically limited by the effect of elevated operating temperatures on photovoltaic efficiency under dynamic conditions. Traditional thermal management strategies limitedly address the non-linear, stochastic, and multi-objective challenges that are inherent to PVT system operation. This paper critically reviews the current application of Artificial Intelligence (AI) as a transformative technology for intelligent thermal management in PVT systems to improve PVT systems’ efficiency.We cover about 130 papers from the last decade, analysing the application of AI paradigms such as Artificial Neural Networks (ANNs), Support Vector Machines (SVM), Deep Reinforcement Learning (DRL) and Physics-Informed Neural Networks (PINNs) to solar PVT systems. The contribution of this work is its focus on thermal management that integrates modern concepts of edge AI, digital twins, and trustworthy AI. It also presents a rigorous comparative analysis of AI against traditional control methods. We also perform analysis through qualitative comparison tables of AI techniques and a visual taxonomy of AI applications. The key research gaps are identified in the study, including the scarcity of standardised validation datasets, the challenge of sim-to-real transfer and the need for a strong and computationally efficient edge deployment. The review then focuses on a strategic research roadmap which advocates for a focus on hybrid physics-AI models, verifiable digital twins, and explainable AI (XAI) to build strong, efficient, and autonomous PVT infrastructures
A vision transformer model-integrated mobile application for early and accurate detection of lumpy skin disease in cattle
No full textLumpy skin disease (LSD) is a highly contagious viral disease of cattle and causes significant economic losses in the livestock industry, around the globe. Early and accurate detection is critical for effective disease management and control in a timely manner. Early-warning digital detection approach such as Convolutional Neural Networks (CNNs) have shown promising results in medical and veterinary diagnostics, Vision Transformers (ViT) remain relatively unexplored in this field. In the proposed research, we used a total of 8000 images of cattle (retrieved from Kaggle) and trained the model to achieve the optimal detection of infections. We used data pre-processing techniques of resizing, normalisation, and augmentation followed by the ViT architecture as a classifier. The model provided excellent performance that showed a 98.12% accuracy, 98.5% precision, 98.5% recall, and a 98.5 F1 score. We have shown that ViT achieves better results in LSD classification compared to multiple approaches that are traditionally used, providing greater accuracy and precision. To encourage adaptation and apply model easier in the field conditions, a mobile application was created and validated on PyTorch Lite and Flutter. Collectively, this powerful approach would change the health management of livestock and allow swifter, and more accurate diagnosis not only to contain the infection but also to counteract its transmission in susceptible livestock
Dichotomies as points of departure : A response to Truscott and Sharwood Smith (2024)
Get PDFWe here respond to a 2024 discussion and commentary article entitled Dangerous dichotomies and misunderstandings in second language research by Truscott and Sharwood Smith (T&SS), who argue that several dichotomies pervade the field of second language acquisition (SLA) that negatively impact progress in the field. T&SS focus on four dichotomies, all of which imply an opposition of generative and usage-based approaches: (i) Cognitive vs. Generative, (ii) Usage-based vs. Generative, (iii) Dynamic vs. Static/Fixed, and (iv) Innatist vs. What? We find T&SS’s specific approach problematic as corrections are overly skewed towards a single side; some imprecisions are simply swapped for others; and at times, crucial developments in both generative and usage-based approaches are ignored. Thus, we – two usage-based and one generative language researcher – combine forces here to offer our perspective. For the ‘dangers’ that T&SS list regarding each of the four dichotomies they discuss, we provide a synopsis of where we agree with T&SS and where we do not; and, based on where we see contemporary generative and usage-based approaches stand with regard to these four dichotomies, we offer an alternative set of statements that we consider more balanced and nuanced than the ‘corrective statements’ initially offered in T&SS (2024)
Exact two-sided confidence sets for a level set in simple linear regression
Get PDFRegression modeling is the workhorse of statistics. It is realized in recent years that one important aim in regression analysis may be the estimation of a level set of the regression function. The published work on this has thus far focused mainly on nonparametric regression, especially on point estimation. In our previous work, we constructed exact upper and lower, but only conservative two-sided, confidence sets for a level set in linear regression. In this paper, exact two-sided confidence sets are constructed in simple linear regression. A simultaneity property of the exact two-sided confidence is also studied. An example is given to illustrate the method
Stochastic dynamic job scheduling with interruptible setup and processing times : An approach based on queueing control
No full textWe consider a stochastic, dynamic job scheduling problem, formulated as a queueing control problem, in which a single server processes jobs of different types that arrive according to independent Poisson processes. The problem is defined on a network, with jobs arriving at designated demand points and waiting in queues to be processed by the server, which travels around the network dynamically and is able to change its course at any time. In the context of machine scheduling, this enables us to consider sequence-dependent, interruptible setup and processing times, with the network structure encoding the amounts of effort needed to switch between different tasks. We formulate the problem as a Markov decision process in which the objective is to minimize long-run average holding costs and prove the existence of a stationary policy under which the system is stable, subject to a condition on the workload of the system. We then propose a class of index-based heuristic policies, show that these possess intuitively appealing structural properties and suggest how to modify these heuristics to ensure scalability to larger problem sizes. Results from extensive numerical experiments are presented in order to show that our heuristic policies perform well against suitable benchmarks
Diversity-oriented route to functional covalent triazine frameworks
Get PDFAn alternative route to synthesise TCNQ-CTF by using trifluoromethanesulfonic (TFMS) acid catalysis is presented, in comparison to a previously reported ZnCl2-catalysed synthesis. The new synthetic route yields a polymer with additional structural diversity compared to the previously reported material. The composition of the framework is rationalised by ‘artificial’ acid-catalysed synthesis of TCNQ-CTF, together with a novel approach to structural feature identification, with a range of alternative structural features appearing that were not present in the previously reported polymer formed by ZnCl2 catalysis. These results will inform the design of new CTF materials with additional functionality and broader applications