It is proposed to use correlation analysis methods to process indirect measurement data when monitoring the incremental сylindrical phase delays referenced to the first cylinder. A method is proposed for restoring the optimal parameters of stratified charge delivery to the combustion chambers of piston-type internal combustion engines. TThe conceptual framework for the development of software and hardware systems incorporating feedback control based on the state of the measurement signal, represented by crankshaft rotational irregularities, has been established. A deterministic mathematical model representing the torque transmission architecture of the powertrain is formulated as a mechanical system with three degrees of freedom, taking into account energy dissipation due to friction. The motions of the masses of the mathematical model are described by a deterministic system of linear differential equations. The parameters of these equations are normalised using the theorems and methods of similarity theory. The Laplace transform under zero initial conditions is applied to solve the resulting system of differential equations. Using the method of determinants and the Mathcad software environment, the information links between the cylinder torques and the signal of uneven rotation of the first crankshaft mass were established. In the Matlab software environment, special points were identified and a simplified representation of the torque transfer functions was obtained as a result of their analysis. A limited Fourier series using Mathcad software approximated the cylinder torques. A computational scheme was developed for simulating deterministic signals characterizing the rotational irregularity of the first crankshaft mass. The additive disturbance superimposed on the measurement signal is modeled as structured white noise with a frequency spectrum constrained to ten harmonic components.
Within the computer modeling framework, the output signal generation utilizes an approach based on the regulation of information pathway lengths in neural network structures to define the gain coefficients corresponding to the aggregated torque amplitudes of individual cylinders. For the first time, an auxiliary algorithm was developed to monitor incremental phase delays of the cylinders relative to the reference (first) cylinder by calculating the mutual correlation function between the rotational irregularity signal of the first crankshaft mass and the torque output of the first cylinder. The software application for calculating the reciprocal correlation function is implemented in the program Mathcad. As a result of the analysis of the mutual correlation function graph, three distinct maxima were identified. The initial peak of the computed mutual correlation function corresponds to the phase associated with the nominal torque generation of the second combustion chamber. The second peak reflects the standard torque phase of the third chamber, while the third peak indicates the reference phase for the first combustion unit. Furthermore, the proportional values of these maxima align with the gain factors assigned to each cylinder's torque in the computational summation scheme. The cross-correlation between the processed measurement signal and the torque signal of the first cylinder was evaluated under conditions of additive stochastic interference. Analysis of the correlation curve demonstrates that a measurement uncertainty of approximately 14% in the rotational non-uniformity of the primary crankshaft mass does not preclude the effective application of correlation analysis techniques for phase shift tracking in the fuel delivery timing across engine cylinders.

Cardiovascular disease (CVD) is a leading cause of death worldwide and hence requires early risk assessment and focused preventative measures. The study describes a novel two-phase hybrid approach that combines machine learning-based CVD risk prediction and personalized lifestyle advice. In the first phase, cardiovascular risk is estimated using ensemble classifier that combines Random Forest Classifier, SVM and LR using metal learner trained on the Heart Disease dataset (1000 record, 14 attributes) has excellent predictive accuracy. In the second phase, optimization framework produces lifestyle suggestions that are safe for health within clinically determined parameters, which are enhanced using a hybrid recommendation system that combines content-based and Cluster-based Outcome Analysis. The suggested approach considerably outperformed a baseline of general lifestyle recommendations in a simulated high-risk cohort, exhibiting an average relative risk reduction of [X] % over a 10-year period as determined by the Framingham Risk Score. The suggested approach is made to be validated in future research using external datasets, simulated patient trials, and physician evaluation in order to guarantee clinical relevance This methodology highlights the promise for precision cardiovascular prevention by providing personalized, data-driven lifestyle recommendations. 

Traditional blood group identification methods, such as serological testing or fingerprint based biometric analysis, require physical contact, specialized equipment and laboratory processing. To remove these boundaries, this study proposes a novel, which is a completely contact -free approach to determine A, B and O blood groups using ocular image analysis. Unlike the previous methods that rely on fingerprint or vein pattern, our technique takes advantage of iris color, conjunctival vasculature, limbal ring intensity, and other eye field features to classify blood group types. A custom dataset of 3,000 eye images was collected from diverse demographics under different lighting conditions. The key features were extracted using hyperspectral imaging and deep learning-based segmentation. We introduce a hybrid multi-modal attention network (HMAN), which integrates transformer-based spatial encoding, convolutional feature extraction, and self-attention mechanisms to enhance classification accuracy. The proposed model obtained 97.1%accuracy, improved ResNet-50 (92.3%) and KushalNet-B4 (94.5%). Ablation studies confirmed that multi-modal feature fusion improves discriminatory capacity for blood group-specific patterns.
This work establishes the first AI-operated, non-invasive blood group detection framework with emergency medical, blood donor screening, and potential applications in biometric diagnostics. Future research will focus on real-time deployment, dataset expansion, and multi-modal physiological feature integration to improve robustness. Our findings represent a major advancement in contact-free medical diagnosis, which paves the way for AI enhance hematological classification.

Timely identification of brain tumors helps improve treatment outcomes and reduces mortality. Accurate and non-invasive diagnostic tools for segmenting and classifying tumor regions in brain MRI scans are crucial for minimizing the need for surgical biopsies. This study builds a deep learning model for tumor segmentation and classification, aiming high accuracy and efficiency. A gaussian bilateral filter is used for noise reduction and to improve MRI image quality. Tumor segmentation is performed using an advanced U-Net model, the Dual Attention AtroDense U-Net (DA-AtroDense U-Net), which integrates dense connections, atrous convolution and attention mechanisms to preserve spatial detail and improve boundary localization. Texture-based radiomic features are subsequently extracted from the segmented tumor  
region using Kirsch Edge Detector (KED) and Gray-Level Co-occurrence Matrix (GLCM) and refined through feature selection to reduce redundancy using the Cat-and-Mouse Optimization (CMO) algorithm. Tumor classification employs an Auction-Optimized hybrid LSTM Network (AOHLN). Evaluated on BraTS 2019 and 2020 datasets, the developed model achieved a Dice Similarity Coefficient of 0.9907 and a Jaccard Index of 0.9816 for segmentation accuracy and an overall accuracy of 98.99% for classification, highlighting its potential as a dependable and non-invasive diagnostic solution.

The fuzzy logic controllers (FLC) gain popularity in ensuring stable and high-performance control of nonlinear industrial plants with no reliable model, where the traditional controllers fail. Their standard expert-based design and simple algorithms that meet the demands for fast execution and economical use of computational resources ease their implementation into programmable logic controllers for wide industrial real-time control applications. This research presents a novel approach to enhancing the performance of FLC systems by compensating for the subjectivity inherent in expert-based design through optimization of the parameters of type-1 (T1) and interval type-2 (IT2) PID FLC membership functions (MF) using genetic algorithms. The approach is demonstrated for controlling the solution level in a carbonization column for soda ash production. Simulations reveal that optimization improves the system performance, measured by a newly introduced overall performance indicator for dynamic accuracy, robustness, and control smoothness, by 48% for the T1 FLC system and 30% for the IT2 FLC system. No improvement is observed in the substitute of T1 MF by IT2 MF for both the empirically designed and the optimised FLC.

Image dehazing is a critical preprocessing step in computer vision, enhancing visibility in degraded conditions. Conventional supervised methods often struggle with generalization and computational efficiency. This paper introduces a self-supervised image dehazing framework leveraging a depth-guided Swin Transformer with hybrid attention. The proposed hybrid attention explicitly integrates CNN-style channel and spatial attention with Swin Transformer window-based self-attention, enabling simultaneous local feature recalibration and global context aggregation. By integrating a pre-trained monocular depth estimation model and a Swin Transformer architecture with shifted window attention, our method efficiently models global context and preserves fine details. Here, depth is used as a relative structural prior rather than a metric quantity, enabling robust guidance without requiring haze-invariant depth estimation. Experimental results on synthetic and real-world benchmarks demonstrate superior performance, with a PSNR of 23.01 dB and SSIM of 0.879 on the RESIDE SOTS-indoor dataset, outperforming classical physics-based dehazing (DCP) and recent self-supervised approaches such as SLAD, achieving a PSNR gain of 2.52 dB over SLAD and 6.39 dB over DCP. Our approach also significantly improves object detection accuracy by 0.15 mAP@0.5 (+32.6%) under hazy conditions, and achieves near real-time inference (≈35 FPS at 256x256 resolution on a single GPU), confirming the practical utility of depth-guided features. Here, we show that our method achieves an SSIM of 0.879 on SOTS-Indoor, indicating strong structural and color fidelity for a self-supervised dehazing framework.

In this research, we established a machine learning–based model to predict the suitability of tsunami evacuation locations in Padang City through the Extreme Gradient Boosting (XGBoost) method. We trained the model on a new synthetic dataset with 5,000 observations with key geospatial and demographic features such as elevation, distance to coastline, suggested evacuation capacity, surrounding population count and site area. The analysis process consisted of preprocessing, feature selection utilizing the XGBoost Classifier, training and cross-validation on each model, and evaluation through regression as well as classification metrics. The XGBoost model performed best (RMSE=0.0642, MAE=0.0418 and Accuracy=93.8%), which was even better than Random Forest, Gradient Boosting Trees and Logistic Regression models. These findings demonstrate that XGBoost can successfully extract complicated spatial–demographic associations with little overfitting. The residual analysis and the actual-vs-predicted plots also reveal good model calibration and stability. A web prototype was also created to visualize the suitability of evacuation and facilitate spatial decision making. Although the model is based on simulated data, it offers an extendible and interpretable framework to be integrated in practical scenarios with field and operational disaster management systems. To the best of our knowledge, this work represents the first use of XGBoost algorithm in Indonesia to classify tsunami evacuation sites and functions as a new tool for disaster preparedness and evacuation plans on the coast.

The rapid growth of IoT ecosystems has intensified the complexity of fog–cloud infrastructures, necessitating adaptive and energy-efficient task offloading strategies. This paper proposes MADRL-MAML, a Multi-Agent Deep Reinforcement Learning framework enhanced with Model-Agnostic Meta-Learning for dynamic fog–cloud resource allocation. The approach integrates curriculum learning, centralized attention-based critics, and KL-divergence regularization to ensure stable convergence and rapid adaptation to unseen workloads. A unified cost-based reward formulation is used, where less negative values indicate better joint optimization of energy, latency, and utilization. MADRL-MAML is benchmarked against six baselines Greedy, Random, Round-Robin, PPO, Federated PPO, and Meta-RL using consistent energy, latency, utilization, and reward metrics. Across these baselines, performance remains similar: energy (3.64–3.71 J), latency (85.4–86.7 ms), and utilization (0.51–0.54). MADRL-MAML achieves substantially better results with a reward of $-21.92 \pm 3.88$, energy 1.16 J, latency 12.80 ms, and utilization 0.39, corresponding to 68\% lower energy and 85\% lower latency than Round-Robin. For unseen workloads characterized by new task sizes, arrival rates, and node heterogeneity, the meta-learned variant (MADRL-MAML-Unseen) achieves a reward of $-6.50 \pm 3.98$, energy 1.14 J, latency 12.76 ms, and utilization 0.73, demonstrating strong zero-shot generalization. Experiments were conducted in a realistic simulated environment with 10 fog and 2 cloud nodes, heterogeneous compute capacities, and Poisson task arrivals. Inference latency remains below 5 ms, confirming real-time applicability. Overall, MADRL-MAML provides a scalable and adaptive solution for energy-efficient and latency-aware orchestration in fog–cloud systems.

Early identification of Parkinson's disease (PD) from MRI remains challenging due to subtle structural alterations and the complexity of brain tissues. To address these challenges, this paper proposes a hierarchical framework termed Hierarchical Severity-Adaptive Diffusion Network, composed of three sequentially connected phases, where the output of each phase serves as input to the next for task-specific optimization. In the first phase, a graph diffusion-based convolutional network is employed to extract anatomical and structural features from multi-modal MRI data, enabling accurate segmentation of PD-relevant regions. Phase two introduces an edge-enhanced slice-aware recurrent network that incorporates Wiener filters and Sobel-based edge enhancement to reduce noise and partial volume effects while capturing structural continuity across adjacent MRI slices. Finally, for severity classification, non-linear severity-adaptive attention network is introduced, which emphasizes discriminative feature deterioration patterns across stages. This model uses Figshare PD dataset and demonstrates superior performance compared to established models like DenseNet121, VGG16, ResNet, MobileNet and Inception-V3, and achieves high accuracy (98.67), precision (0.99), recall (0.98), and F1 score (0.99), indicating its potential as an AI-assisted tool for PD severity assessment using MRI.

This research explores the automated leaf-based identification of medicinal plants, utilizing machine learning and deep learning techniques to address the crucial need for efficient plant classification. Driven by the vast potential of medicinal plants in pharmaceutical development and healthcare, the study aims to surpass the limitations of existing methodologies through thorough experimentation and comparative analysis. The primary goal is to develop a robust and automated solution for classifying medicinal plants based on leaf morphology. The methodology encompasses acquiring diverse datasets. Specifically, Set 1 data is processed by applying resizing, rescaling, saturation adjustment, and noise removal, while Set 2 data is processed by applying resizing, rescaling, saturation adjustment, noise removal, and PCA (Principal Component Analysis). The proposed algorithms include Support Vector Machines (SVM), Convolutional Neural Networks (CNNs), YOLOv8, Vision Transformer (ViT), ResNet, and Artificial Neural Networks (ANN). The study evaluates the efficacy and effectiveness of each algorithm in plant classification using metrics such as accuracy, recall, precision, and F1 score. Notably, the ResNet model achieved 93.8% and 94.8% accuracy in Set 1 and Set 2, respectively. The SVM model demonstrated 56.5% and 56.6% accuracy in Set 1 and Set 2, while the Vision Transformer (ViT) model achieved 84.9% and 74.4% accuracy in Set 1 and Set 2, respectively. The CNN model showcased high accuracy at 96.7% and 94.8% in Set 1 and Set 2, followed closely by the ANN model with 96.7% and 96.6% accuracy. Lastly, the YOLOv8 model achieved 96.0% and 95.1% accuracy in Set 1 and Set 2, respectively. The comparative analysis identifies CNN and ANN as the top-performing algorithms. This research significantly contributes to the advancement of medicinal plant identification, pharmaceutical research, and environmental conservation efforts, emphasizing the potential of deep learning techniques in addressing complex classification tasks.