The stress-strain behavior of carbon fiber reinforced polymer matrix composites (CFRPs) is highly influenced by thermal expansion mismatches, particularly due to their anisotropic nature and directional mechanical properties. While prior research has explored thermal and mechanical loading separately, a critical gap remains in understanding their combined effects concerning fiber orientation and load application. This study integrates finite element modeling (FEM) with optimized coefficient of thermal expansion (CTE) selection to improve stress-strain prediction accuracy under coupled thermal-mechanical loading for various fiber orientation. Experimental tensile tests at-53 degrees C, 82 degrees C, and room temperature reveal that improper CTE selection significantly impacts transverse loading, causing premature matrix cracking and interfacial debonding. The proposed FEM framework captures thermally induced residual stresses, improving predictive accuracy and reinforcing the need for thermo-mechanical coupling in aerospace applications to ensure long-term material reliability.
Bottom plate leakage is a critical failure mode in storage tank engineering, often causing environmental contamination and high remediation costs. To improve structural integrity and leak detection, double-layer steel bottom systems have been widely adopted and standardized in API 650 and EN 14015. This study evaluates five typical configurations (Structures A-E) using finite element analysis with both 2D and 3D models under hydrostatic loading. Structure D shows the most uniform stress distribution, while Structure E achieves minimal deformation with greater complexity. Parametric studies investigate the effect of rubber pad thickness and elastic modulus on stress and deformation, revealing limited influence on overall performance. A modulus-material mapping framework is established, correlating elastic modulus to commercially available elastomers such as EPDM, polyurethane, and HDPE. Polyurethane materials with elastic moduli between 10-50 MPa are identified as offering the best balance between mechanical performance and economic feasibility. This research provides theoretical and practical insights for optimizing double-layer tank bottom structures, supporting better material selection, safer designs, and more efficient long-term operation of liquid storage systems.
A key goal of engineers working on industrial fluid management systems is to balance cost and performance. This research presents a methodology for redesign of a liquid dispersion unit using value engineering (VE) concepts and simulation-based analysis. The approach identifies high-cost, low-value components through Function-Cost-Worth Analysis (FCWA) and evaluates their impact with the Function Analysis System Technique (FAST) and evaluation matrices. Results showed that the nozzle subsystem was the main limitation, as it did not cover the spray area well and fluid dynamics were unsatisfactory. To address this, several nozzle shapes and materials were examined during the creative phase. A complete cone nozzle proved the best design. Computational Fluid Dynamics (CFD) confirmed improved flow properties, while Finite Element Analysis (FEA) showed structural adequacy under operational pressures. Polypropylene was selected as the best material since it is secure and significantly less costly compared to metals like gunmetal and stainless steel. The changes increased system efficiency from 63.75% to 75.25%, an 11.5% improvement, while overall cost decreased by 2.52%. This demonstrates that integrating VE with CAD-based simulations can generate innovative, scalable designs for fluid-based industrial systems.
Bio-Inspired Structures have become a significant research direction for developing sustainable Structural components with lightweight yet damage tolerant, and energy efficient designs. Among these Structures, one of the most prominent ones is the Bouligand-type Helicoidal structures which are characterized by their linear and gradual rotation of layers between each layer. They've shown a high resistance to fracture and improvised energy dissipation capabilities. while significant number of studies were conducted to examine their high strain-rate impact behavior, their response to gradual loading are has not been full studied. To address this gap, the present work involves in performing an investigation of Bouligand-type structures under quasi static loading using Ansys. In this study, four such Bio-inspired structures were studied out of those four, three of them are helicoidal structures (Bouligand-type) with different angle of layer rotations and the other one is Honeycomb Structure. All these structures are assigned epoxy-carbon (395 GPa) composite material to ensure consistent mechanical assumptions across all the cases. All of those structures are designed and sandwiched between two plates. A Controlled displacement ranging from 0-10 mm was applied to the top surface which is generally the plate in a quasi-static Approach, while the bottom plate is fixed. In addition to compression, a displacement-controlled flexural analysis was conducted to evaluate the structural performance under bending action and to represent more realistic service loading conditions. Frictional effects were neglected to maintain computational stability and to solely focus on the structural response.
This research presents a machine learning-based approach for monitoring an industrial battery production system using a Naive Bayesian Network, a probabilistic model widely recognized for its ability to handle uncertainty. The proposed framework infers system states from observed operational conditions and event data, providing predictive insights into machine behavior. Real-world production data were employed to train and validate the model, ensuring both accuracy and practical applicability. Through probabilistic inference, the model anticipates potential failures or abnormal behaviors, enabling timely maintenance interventions and minimizing downtime. Evaluation results demonstrate that the Naive Bayesian Network offers a robust and interpretable solution for industrial monitoring, with strong potential to enhance predictive maintenance strategies and improve the overall reliability and efficiency of battery manufacturing operations.
Diabetes is a chronic illness with high morbidity and mortality that influences the quality of life considerably around the globe and thus early and correct prediction is crucial in the effective management and treatment. Nonetheless, the clinical information used is problematic because it is difficult to predict diabetes development in patients, given the complexity and variability of the data. This paper proposes a deep learning-based model with a Long Short-Term Memory (LSTM) recurrent neural network and improved preprocessing and feature selection algorithms. First, Z-score normalization is used to standardize the data, enhancing consistency and identifying abnormalities. Then, to achieve the best feature selection, the Grey Wolf Optimization (GWO) is used to improve predictive performance by identifying the most relevant clinical attributes without falling into local optima. Lastly, the LSTM-RNN model is applied to extract temporal dependencies and latent patterns in the data to correctly classify the data. Through experimentation, it has been shown that the proposed approach clearly exceeds conventional techniques based on all available measures of performance: accuracy; precision; recall; F1 score; and computational efficiency. As indicated by these results, this LSTM-RNN-GWO model shows promise as a valuable resource in the area of predictive analytics related to diabetes care, providing great benefit to patients through its use in early identification of diabetes and subsequent enhancement of their clinical experience.
Breast cancer recurrence is one of the most significant medical concern, and accurate recurrence models can assist in early intervention and treatment planning. Breast cancer recurrent remains as one of the most critical concern for patients prognosis and treatment planning. Accuracy Predicting individual recurrence risk is crucial for the development of precise therapy, specialy for those patients with high-risk profiles. In the study proposes a hybrid machine learning approach that uses the computational modeling and the medical information to predict the recurrence of breast cancer in a patient. The dataset contains the medical and patient information like the age, tumor size, lymph node involvement, malignancy degree, location, irradiation status and recurrence class. This proposed approach begins with the process of data processing, handling the missing data values, features normalization and encoding of categorical variable into numerical format. The dataset is divided into two parts the training set and the testing set and the two selected models' random forest and logistic regression models are trained independently. The predictions form both the model is stacked and a logistic regression meta-model is trained on these combined predictions. The evaluation of the model was conducted using the metrics such as accuracy, precision, recall, and F1 score. The designed hybrid model was able to achieve the accuracy of 97.66% with the precision, recall and F1 score all reaching around 98.15%. This study highlights the potential of hybrid machine learning techniques, improving the accuracy and reliability of machine learning models for breast cancer recurrence prediction. This development model can serve as a valuable tool for the medical industry to support decision making and assist in personalized treatment decisions, offering early detection of recurrence. This can enhance the treatment of a patient by supporting early detection and patients' outcomes through targeted therapy.
Pharmaceutical industry works under the strict Good Manufacturing Practice (GMP) conditions which require strong systems to guarantee the quality of the product, integrity of data and adherence to regulations. Manufacturing Execution Systems (MES) have become one of such key tools in the realization of such goals. The purpose of the study is to assess the application and functioning of PAS X MES among the GMP-regulated manufacturing facilities, in particular, its operational effectiveness, compliance benefits, as well as difficulties faced. Multi-site observational analysis was used where site survey, system audit data and key stakeholder interviews were used. Pre-and post-implementation performance measures were measured quantitatively. The outcomes have shown 25 percent decrease in the batch cycle time, increasing Right First Time rates to 95 percent and a drop in process deviations by 66.7 percent. The metrics on compliance were highly improved where there was a 100 percent compliance with the 21 CFR Part 11 standard on the use of electronic signatures and the audit score has improved by 38.5 percent. Additionally, the time taken in review by QA per batch decreased by 62.5 percent making the processes of release of batches faster. Along with these improvements, issues that included spending more time with the validation in the first place and the reluctance of the users in the beginning made clear the necessity of a proper management of the change. Comparative study with other MES systems determined that the implementation speed and user satisfaction characterized PAS X with competitive advantage. This paper highlights PAS X MES as a game-changer to GMP-compliant pharmaceutical manufacturing, which can bring organizational operational agility and digital maturity to Pharma 4.0 movements.
Data anonymization in healthcare is essential for protecting sensitive patient information while enabling secure usage for research, analytics, and AI-driven clinical decision-making. In this study, the MIMIC-III-Deep Reinforcement Learning dataset was used, which contains comprehensive electronic health records (EHRs) of ICU patients. Data preprocessing was performed using Min-Max Normalization to scale numerical features and ensure consistency. Anonymization techniques such as pseudonymization, generalization, suppression, data masking, and statistical methods like k-anonymity, l-diversity, and t-closeness were applied to safeguard patient privacy. The anonymized dataset was then utilized for predictive modelling using AI techniques including Random Forest and LSTM. Results demonstrated that privacy was maintained with 0% PII leakage, while predictive accuracy remained high, achieving accuracy of 94.6%, precision of 93.8%, recall of 92.5%, and F1-score of 93.1%. This study highlights that effective data anonymization ensures compliance with HIPAA and GDPR while retaining the utility of healthcare data for advanced analytics and AI applications.
Chronic Kidney Disease (CKD) is among the most significant global health concerns, particularly in terms of its insidiousness during the first stage of its development and gradual devastation throughout the years. There is a prospect of utilizing Electronic Health Records (EHRs) to improve the outcome due to the ability to address problems at an early stage to deliver the most efficient intervention. The paper presents an intelligent predictive analytics system of healthcare in Abu Dhabi healthcare systems that is built on the EHR data collected. The pipeline of the framework is systematic and it entails data preprocessing, feature extraction and classification. The preprocessing phase is assigned to aligning the data, its coherence, and the removal of redundancies and the handling of missing values across all the EHR datasets. The step is relevant due to the heterogeneous nature of clinical information being rather complex. In summarizing the data, Principal Component Analysis (PCA) is applied to extract the features by subjecting the data to the process to compress the data and retain the most clinical information. This improves the computational and model efficiency and performance by removing noise and redundancy. It is then inputted into the constructed Long Short-Term Memory (LSTM) network due to its learning capabilities which give long-range dependencies and temporal patterns of sequential patient information. Precision, recall and F1-score, are also used to test the effectiveness of the model by determining whether the model is effective in the proper identification of CKD cases. The findings show that LSTM model is better than the traditional classifiers it is more predictive and robust. As highlighted in this paper, advanced deep learning methods might be used on EHR data to aid in the prompt identification of CKD and enhance the clinical decision-making process. The suggested framework is flexible and can be extended and provide useful information on how the framework can be applied in real-life healthcare.
Leiomyosarcoma is a rare type of cancer that spreads to various parts of the body to create an aggressive form of complex tissue sarcoma disease. Artificial Intelligence (AI) powered technologies play a vital role in screening medical images to identify sarcoma types of diseases for early diagnosis and treatment to avoid cancer risks. In the preliminary stages, the machine learning and deep learning models potentially impact the identification of cancer levels; due to the invasive point of image degradation and feature inconsistency levels, the precision level attains low accuracy, leading to higher false negatives. To solve these problems, to propose an Optimal Particle Swarm Intelligence Technique (OPSIT) for feature selection with Long Short-Term Memory Gated Recurrent Neural Network (LSTM-GRNN) to identify the disease effectively. Initially, a bilateral wavelet filter (BWM) is carried out preprocessing to normalize the feature scaling and improve the image scalar margins. Then, the Linear Reiterative Clustering (LRC) algorithm is applied to segment the non-invasive point of the disease scalar region to split the cancer cells. Further, to scale, the active disease margins in cancer cell features are evaluated with OPSIT to reduce the non-relation feature in the segmented image region. Finally, the LSTM-GRNN algorithm is applied to train the scaled entity of the cancer image region, with Actual threshold margins to identify the disease region accurately. The proposed system increases the proper positive actual scaling region of the cancer region to increase precision rate and attain high accuracy, sensitivity, specificity, and ROC performance compared to the other systems.
Numerous disorders related to lifestyle choices and environmental factors are prevalent among humans today. Predicting and detecting these diseases early on is essential to halting their spread and severity. For physicians, accurately diagnosing illnesses can be challenging. Specifically, one of the key origins of morbidity and death from non-communicable diseases that impact 10-15% of the global population is chronic kidney disease, or CKD. Still, making medical predictions is a difficult and complex undertaking. Our proposed system uses powerful machine learning algorithms to detect and predict people with prevalent chronic conditions. These methods can enhance classifiers' ability to reliably identify chronic diseases. The dataset collected from Kaggle is a chronic kidney disease dataset, comprising 25 features. The first step is preprocessing and normalization of the dataset. PCA extracts the features of chronic disease. The k-nearest neighbour (KNN) is a feature selection method used to select features. A CNN (convolutional neural network)-GRU (gated recurrent unit) classification algorithm is used to predict disease from the dataset. The predicted result is binary, like "CKD" or "NOT CKD", The classification algorithm efficiently evaluates performance metrics, including precision, accuracy, recall, and an F1 score of 1.0.
Neuroblastoma is the most common extracranial solid malignancy in children. It is possible to estimate the cancer's unpredictable biological activity based on the patient's age, genetic makeup, the biology of the cancer, and the extent of disease at diagnosis. Machine learning algorithms have the potential to enhance the precision and efficacy of cancer diagnosis, individualized therapy selection, and long-term outcome prediction. A subset of machine learning known as Artificial Intelligence (AI) is capable of spotting patterns in data and acting without the need for special programming to accomplish predetermined objectives. The patient populations most likely to benefit from advanced imaging tests may be enriched, high-risk populations can be identified, and individualized screening tests can be prescribed with the aid of machine learning technologies. Modern computational tools are becoming more and more crucial in pediatric oncology because of their invasive nature and the requirement for an early and precise diagnosis. Convolutional Neural Networks (CNNs) and Gated Recurrent Units (GRUs) are two methods that can be used to increase the precision and predictability of detection. The Grey-Level Cooccurrence Matrix (GLCM) extracts the features as energy, entropy, dissimilarity, homogeneity, and contrast. The CNN-GRU to produce the results as precision, recall, accuracy, and F1-score 93%, 80%, 95%, and 83%.
Breast cancer is a major health issue, and effective treatment depends on a prompt diagnosis. Particularly mammography is important in the detection of breast cancer. Deep learning algorithms have shown promise in analyzing medical images, but their performance heavily relies on large labeled datasets, which are often limited in the context of breast cancer. In spite of the lack of labeled data, this study suggests a unique method called Cross-Dimensional Transfer Learning to increase the precision of cancer identification using deep learning. The method utilizes multiple imaging modalities, such as mammography and ultrasound, to leverage the complementary information and transfer knowledge learned from one modality to enhance classification performance on another. The proposed work consists of following three phases: Pretraining on Diverse Data, Modality-Specific Fine-Tuning and Cross-Dimensional Transfer Learning. A deep learning model is pretrained on a diverse dataset that includes breast cancer images from different modalities. This phase enables the model to learn general features and representations applicable across various imaging modalities. After pretraining, the model is perfected using labeled data specific to each modality. This process enables the model to adapt its learned features to the exclusive features of each imaging modality, improving its ability to capture modality-specific patterns related to breast cancer. Once modality-specific fine-tuning is complete, knowledge acquired from one modality is transferred to another by leveraging shared representations between the imaging modalities. This transfer of knowledge enhances the classification performance on the target modality, particularly when labeled data is limited for that modality.
The biomedical industry uses graph mining to store a variety of data. It features a feature that makes a lot of info accessible. However, the majority of individuals are mainly concerned with learning about illnesses and medications. Creating drug models and novel chemical molecules in the medical industry is called relational medicine. Patients may have adverse reactions to the medication due to the dissimilarity of the compound's molecules. It suggests that complex molecular characteristics don't affect the dataset's classification and are independent of one another. We aim to increase the effectiveness of graph mining-based drug selection by analyzing the success rate of drug selection using Fuzzy Multilayer Neural Perception (FMNP). Marginal Subset Clustering Features (MSCF), the input used to generate the chosen classes, are processed utilizing these features. The system first preprocessed all patient features and suggested medication molecule compounds to create a consolidated dataset. Distance vector features linked to edge weights are used in feature selection to create relationship patterns. The features are chosen based on the estimated Relational Drug Combination Weight (RDCW). Additionally, the implementation updates the logic rules and gives the neural classifier predictions for feature weights. Using a neural classifier and iterative logic rules, FMNP forecasts training outcomes. The classifier continuously predicts and suggests chemical molecules to lower the possibility of adversative effects.
This study proposed an Attention Algorithm Based-Random Forest model (AAB-RF) To enhance the precision of the ML model, dataset 1 focuses on cardiac disease, including metrics like age, cholesterol, blood pressure, and other cardiovascular factors. The data set 2 aims at the mitral valve issues, especially focused on detecting the prosthetic valve anomalies. The two datasets are analyzed using machine learning classifiers, including K-Nearest Neighbors, Random Forest, Naive Bayes, Logistic Regression and advanced methods like Convolutional Neural Networks and a VGG-based framework. When analyzing the data set 1, the proposed AAB-RF model achieves the classification accuracy of 93% performs better than the other models like Naive Bayes of 88.52% and Support Vector Machine of 89% accuracy. Likewise, for data set 2, the proposed AAB-RF model reached the remarkable accuracy of 99.40% performs superior than CNN and VGG achieved an accuracy of 97% and 84% respectively whereas closely matching with the Vafaeezadeh et al. (2021) model's accuracy of 99.00%. The major advantage of integrating attention mechanism with the Random forest model enhances the feature selection and decision making, especially in datasets having the sophisticated interdependent nature. This study showcasing the AAB-RF model's efficiency in managing the multiple datasets which enables the robust and effective outcomes. The research outcomes shows its effectiveness which guides the physicians in diagnosing the heart disease and interpreting the mitral valve features with high accuracy and reliability.
The integration of AI-driven computation, real-time data streams, and image processing is reshaping traffic management and urban logistics optimization. This research builds on the NarrQuest system-a globally pioneering narrative-computational methodology formalized through a five-article methodological canon and fifty published monographs-to introduce a first-of-its-kind logbook-based optimization framework. In this approach, personal journey narratives are encoded into structured search heuristics, transforming subjective records into formal routing constraints. Classical combinatorial optimization models, from Travelling Salesman Problems (TSP) to Vehicle Routing Problems (VRP), Integer Programming (IP), Scheduling, and Queueing Theory, are reformulated within this narrative optimization paradigm. Using multi-campus delivery datasets, exhaustive enumeration resolves small-scale TSP instances, while VRP formulations incorporate multi-agent constraints such as vehicle capacity and time windows. Integer Programming enhances modeling flexibility under contextual constraints, and Scheduling with Queueing theory stabilizes dynamic system performance. Large-scale complexity is addressed through AI-powered metaheuristics, particularly Genetic Algorithms for adaptive routing. Realtime image processing-including computer vision traffic sensing and behavior-informed demand forecasting-further strengthens responsive decision-making. Game-theoretic models capture strategic interaction within dynamic logistics ecosystems. This work positions NarrQuest not merely as a theoretical contribution, but as the first narrative-driven computational architecture with demonstrated capacity to meet other indexed algorithmic publication standards, bridging lived experience and intelligent logistics computation.
The applications of composite materials have been increasing significantly in recent decades due to their superior mechanical properties and versatility. The major effect limiting the use of composite materials is the lack of understanding of their response and their structural integrity under dynamic loads. Among the prominent damage mechanisms, the debonding under dynamic loading is a well-recognized failure mode for laminated composites. Up to date, the impact of the significant parameters on the delamination is thoroughly examined in this study with primary focus on the hemispheric indenter diameter and the characteristics of the exerted load applied at constant energy levels. The damage morphology has been carefully investigated using X-ray computed tomography, quantifying the shape and size variation of delamination areas across plies. The experimental observations have been incorporated into the finite element modeling, carried out in ABAQUS, by means of cohesive elements, which allow for the setting of a failure criterion. The main delamination area has been confirmed to be localized on the tension side of the laminate, where the most bending stress is sustained. Moreover, the angular difference between adjacent plies that articulates the distribution of the interlaminar stresses has to be taken into consideration, since it has a great impact on the extent of delamination. It is concluded that the initiation of delamination can be detected using a delamination threshold load based on the quasi-static load-displacement curve. These results illustrate the importance of the indenter radius to thickness ratio as a governing parameter in the structural response of composite plates, aiding in the development of more accurate predictive models for damage assessment.
Functionally graded materials (FGMs) are advanced materials with varying material properties directionally. The material properties, such as elastic modulus, density, thermal conductivity, and thermal expansion coefficient, improve by combining different materials. The structure thus offers a better strength-to-weight ratio, thermal resistance, and durability in critical harsh environments. In the present study, material properties follow a power law for material gradation along the thickness of the cylinder. Navier's approach is followed to solve the second-order governing differential equations with the assumption of a plane stress condition to eliminate the complexity of differential equations, and MATLAB was used to solve stress and deformation variation analytically and for visualization. This research aims to provide an exact solution to stress and deformation for different real-life complex loading conditions with material non-homogeneity. It is essential to analyze various loading conditions to maximize the structure's lifecycle, minimizing the stresses induced. This analysis provides valuable insight to enhance the performance, reliability, and integrity of structures in the fields of aerospace, defense, mechanical, and civil. This research also provides a prominent connection for practical application to designing/analysis.
In the present study, free vibration analysis of an FG plate has been performed by employing trigonometric shear deformation plate theory. The selection of an appropriate homogenization model is important as it could significantly influence the material behavior. Therefore, well-known micromechanical models such as Voigt, Reuss, and representative volume element method have been studied and their results are compared. The effect of various boundary conditions was also seen by changing the boundary conditions as SSSS, CCCC, CSCS, and FCFC. The mechanical properties change uni-directionally across the thickness according to a simple power law. Hamilton's principle is applied to derive the governing equations of motion, and Navier-type analytical solutions are formulated for vibration analysis. The study examined the impact of the power-law index, length-to-thickness ratio, micromechanical models, and boundary conditions on the natural frequencies of the FG plate.
