Aquaculture is developing in many countries to meet the rising protein demand associated with a rapidly growing human population. However, intensification of production brings with it challenges such as deteriorating water quality and increasing disease outbreaks. These conditions lead to an increased risk of infectious diseases and led to higher mortality rates, which causes economic losses. Lactococcosis, particularly that caused by the Gram-positive bacterium Lactococcus garvieae, is one of the most common of these problems and can cause significant losses in various fish species. The intensive use of antibiotics for control purposes creates additional risks in terms of antimicrobial resistance, environmental pollution, and food safety. Therefore, natural feed additives have gained importance in recent years. Certain additives that support immune responses and increase disease resistance in fish have become prominent. Among these, phytobiotics, probiotics, prebiotics, and synbiotics are the most well-known and effective. Their widespread availability, lower cost, and environmental safety make these additives considered an alternative approach to aquaculture. Studies show that these additives strengthen both innate and adaptive immune responses, reduce infection severity, and reduce mortality associated with L. garvieae infections. However, there are still gaps in knowledge regarding how these substances regulate mechanisms such as the immune system, inflammatory processes, antioxidant defenses, and interactions with pathogens. This review aims to clarify these mechanisms by bringing together scientific data obtained in recent years. It also discusses how the information obtained can contribute to the development of safer feed additive strategies and the development of new vaccine approaches. This aims to support the establishment of a more sustainable production structure in the aquaculture sector.
The miniaturization of devices alongside advances in thermal management technologies necessitates the generalization of heat conduction and thermal elastic coupling to faithfully represent material responses at ultrashort temporal scales. Motivated by viscoelastic mechanical analogies, this work develops an analytical framework for investigating vibrational behavior in an orthotropic, size-dependent piezo-thermoelastic substrate featuring voids, modeled within the Modified Lord–Shulman (MLS) thermoelasticity theory augmented by fractional derivatives. Employing the Klein–Gordon nonlocal elasticity formulation, the governing equations of motion are rigorously derived. The normal mode method facilitates the examination of coupled thermo–electro-mechanical excitation phenomena. Emphasis is placed on a corrugated interface contiguous to a vacuum, where comprehensive boundary conditions encompassing thermal, electrical, mechanical, and stress equilibria are imposed to determine fundamental field variables. The study systematically evaluates the influence of pivotal parameters, including temporal evolution, nonlocality characteristics, and spatial coordinates, on the thermomechanical and electrical responses, with outcomes substantiated through detailed graphical representations. Although previous investigations have addressed vibrations in porous piezo-thermoelastic media under varying theoretical constructs, the current research uniquely elucidates the dynamic response of a size-dependent porous piezo-thermoelastic medium with a corrugated surface within the fractional-order modified Lord–Shulman framework, marking a significant advancement in the modeling of smart microstructured materials.
Accurate prediction of Remaining Useful Life (RUL) is crucial in the aviation sector to maintain operational safety, minimize unforeseen failures, and facilitate effective predictive maintenance. Deep learning methodologies, including Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, have demonstrated encouraging results in RUL estimation; however, they are hindered by notable limitations. These models frequently learn misleading correlations instead of the genuine degradation patterns of aircraft engines and yield deterministic predictions without assessing uncertainty, which diminishes their dependability in safety-sensitive contexts. To tackle these issues, this paper introduces an innovative framework that combines causal discovery with a Temporal Fusion Transformer (TFT) architecture for dependable and interpretable RUL prediction. The Peter-Clark Momentary Conditional Independence (PCMCI) algorithm is utilized to pinpoint causal degradation features and remove irrelevant sensor noise, thus enhancing feature significance and mitigating overfitting. The identified causal features are subsequently processed through a TFT-based model that incorporates multi-head self-attention mechanisms to adeptly capture both short-term and long-term temporal degradation dependencies. Furthermore, quantile loss optimization is applied to produce probabilistic RUL predictions, allowing for uncertainty quantification via the development of a Cone of Uncertainty. Experimental assessments on the NASA C-MAPSS FD001 dataset reveal that the proposed approach achieves an RMSE of 17.06 and an empirical P10-P90 coverage of 88 percent, surpassing traditional CNN- and LSTM-based baseline models in both prediction accuracy and reliability.
This study presents a multi-scale computational framework to evaluate the Mode-I interlaminar fracture toughness of carbon nanotube (CNT)–enhanced glass fiber reinforced polymer (GFRP) composites. A two-step Representative Volume Element (RVE)-based homogenization approach is employed to predict the effective elastic properties of the CNT-modified epoxy and the resulting laminated composite. These properties are integrated into a macro-scale finite element model of a Double Cantilever Beam (DCB) specimen, where delamination behavior is simulated using the Virtual Crack Closure Technique (VCCT). VCCT enables direct and efficient computation of strain energy release rates and reliable prediction of crack initiation and propagation. The results show a significant increase in critical load and fracture toughness with increasing CNT content up to an optimal volume fraction of 2%, beyond which agglomeration effects degrade performance. Scanning Electron Microscopy (SEM) observations confirm CNT-induced toughening mechanisms such as crack deflection, pull-out, and bridging. The proposed framework provides a reliable virtual testing approach for optimizing the interlaminar fracture performance of CNT-reinforced laminated composites.
Diabetic Retinopathy (DR) is a progressive microvascular condition resulting from diabetes. Conventional early detection techniques - fundus photography, optical coherence tomography - suffer from inaccuracy and technological complexity. This paper proposes an enhanced deep learning architecture integrating a Bilinear Double-Order Filter (BDOF), Vision Transformer (ViT) with a Wasserstein Deep Convolutional Generative Adversarial Network (WDCGAN), optimized with the Giza Pyramid Construction Algorithm (GPCA). BDOF is used for pre-processing to enhance retinal structures and suppress noise. WDCGAN uses a selected subset of fundus images from minority DR classes to form high-quality synthetic images, which are merged with real fundus images to form a balanced training dataset. The Vision Transformer exploits self-attention mechanisms to capture global contextual information and improve discriminative feature learning for DR classification. Furthermore, GPCA is employed to optimize the weight parameters of the ViT–WDCGAN architecture, enhancing convergence and classification accuracy. Experimental assessments conducted on openly available fundus image datasets - APTOS-2019 and Diabetic Retinopathy Detection (DRD), prove that the put forward approach outdoes existing models in accuracy, sensitivity, specificity, and area under the curve. ViT-WDCGAN-BDOF-GPCA yielded an overall accuracy of 98.20% and 98.62% with APTOS-2019 and DRD datasets respectively.
