Madhav Institute of Technology & Science Gwalior

This study presents the fabrication and comprehensive tribological assessment of Al6061-based hybrid composites reinforced with Titanium diboride (TiB2) and cow dung ash (CDA) using the stir casting technique. The wear behavior of TiB2-CDA/Al6061 composites was systematically analyzed under dry sliding conditions utilizing a pin-on-disc setup. The study investigates the effects of key parameters, including reinforcement percentage (R), applied load (L), sliding velocity (V), and sliding distance (D), on wear loss and the coefficient of friction (COF) through a full factorial experimental design. Additionally, scanning electron microscopy (SEM) was employed to examine dominant wear mechanisms under extreme wear conditions, revealing adhesion, abrasion, oxidation, and delamination as primary degradation processes. Furthermore, machine learning techniques, including Random Forest (RF), Support Vector Machines (SVM), Gaussian Process Regression (GPR), and Gradient Boosted Trees (GBTA), were leveraged to develop predictive models for wear loss and COF. The models were trained and validated using experimental data, demonstrating the efficacy of machine learning in accurately predicting tribological performance while minimizing extensive experimental trials. Among the models, GPR exhibited the highest predictive accuracy, surpassing other algorithms in forecasting wear behaviour.

This paper focuses on the utilization of hesitant and intuitionistic fuzzy sets (HFS & IFS) in a computational intelligent approach, mainly for decision modelling under complex vague surroundings. Indeed, classical fuzzy sets can be powerful in some applications, but they may not describe the broad range of epistemic imprecision and non-stationary uncertainty involved in making decisions at any given time. HFS & IFS overcome it by providing the ability to quantify degrees of hesitation and membership uncertainty. HFS allows a decision maker to express the multiple reference degrees associated with an option and also show uncertainty about how to assign that precise level of degree or real value. On the other hand, IFS allows for more possibilities of representing ambiguity. In this research, we incorporate the proposed model into a computational intelligent approach for improving decision-making, particularly in the context of multi-criteria decision analysis (MCDA), resource allocation and task completion parameters. The results show that the proposed HFS-IFS method presents improved performance in terms of accuracy and uncertainty treatment across a series of case studies.

Elephant migration is essential for preserving biodiversity, but accurately predicting their movement patterns is challenging due to the influence of environmental, human, and ecological factors. This research introduces a machine learning-based approach to predict elephant migration routes between Bandipur National Park and Wayanad Wildlife Sanctuary. The study uses 34 months of historical data, including variables such as temperature, humidity, air quality, vegetation health, and water availability. The dataset underwent thorough preprocessing, including outlier handling, feature selection, and data balancing using SMOTE. Several machine learning models were tested, with Logistic Regression yielding the best results—achieving 94% accuracy—surpassing models like Random Forests, Decision Trees, Naive Bayes, Support Vector Machines, and Neural Networks. The analysis identified important environmental factors, such as seasonal water presence and temperature changes, as key triggers for migration. Additionally, hyperparameter tuning helped refine the models further. The findings show that predictive modeling can aid in wildlife conservation, minimize conflicts between humans and elephants, and inform environmental policy. Future developments will focus on integrating real-time tracking and expanding the range of ecological indicators to improve the model’s effectiveness in changing conditions.

The plastic deformation behavior of selective laser melting (SLM) 3D printed SS316L steel has been analyzed at the temperature range 25- 1000℃ (25 (room temperature), 200, 400, 600, 800 and 1000℃) and strain rate range 10-3-103s-1 (10-3, 10-2, 10-1, 100, 101, 102 and 103 s-1) under compressive loading environments. The flow stress vs. plastic strain results revealed that the flow stress was reduced 136.64% from room temperature to 1000℃ at 10-3s-1. Further, the flow stress was decreased 102.86% from room temperature to 1000℃ at 103s-1. The flow stress was increased 46.63% from 10-3s-1 to 103s-1 at room temperature. Moreover, the flow stress was increased 95.07% from 10-3s-1 to 103s-1 at 1000℃. The temperature and strain rate effect on strain rate sensitivity (m) has been observed for SLM 3D printed SS316L steel. Based on strain rate sensitivity (m), the power dissipation efficiency ([Formula: see text]) and instability dimensionless parameter ([Formula: see text]) map plot contours have been investigated under various hot working parameters for SLM 3D printed SS316L steel. Further, hot working processing maps have been generated by superimposing instability dimensionless parameters ([Formula: see text]) map on the power dissipation efficiency ([Formula: see text]) map for SLM 3D printed SS316L steel. The processing map was further related with investigated material microstructure to identify the hot processing safe and unsafe zone for SLM 3D printed SS316L. The unsafe instability region occurred at the low strain rate range (10-2 - 10-1 s-1), high strain rate range (102-103 s-1) and temperature range (200-400℃, and 800 - 100℃) for 0.02, 0.04, 0.06, 0.08 and 0.10 strain. Further, the remaining area was useful for hot workability. The Vicker's hardness revealed that the hardness was decreased with 3.87%, 12.55%, 22.01%, 32.35%, and 43.70% at 2000C, 4000C, 6000C, 8000C and 10000C respectively with respect to room temperature hardness.

Natural convection, driven by buoyancy, is utilized for the heat transport in various applications including thermal exchangers, cooling of heat sources, solar collectors, geothermal power systems, electronic devices, microelectronics, and nuclear industries. This research focuses on the free convection of a hybrid nanoliquid containing Ag–MgO nanoparticles in an enclosure having partially active borders. The hybrid nanosuspension utilized is a mixture of MgO and Ag nanoparticles in equal proportions, suspended in water as the base liquid. The square enclosure is subject to the Lorentz force impact. The study examines two cases. In Case 1, the left wall experiences heat dissipation via a heat sink at a fixed temperature T c , whilst the right wall is partly affected by the active chamber borders with a heater at temperature T h (where T h > T c ). The rest sections of vertical borders are adiabatic. In addition, the cavity is thermally insulated on both the upper and lower surfaces. In Case 2, the chamber's vertical sides are heated to a certain extent ( T h ), whereas the bottom wall is somewhat cold ( T c ) and has some level of activity. The remaining inactive sections of the cavity are adiabatic. The control flow equations were resolved with the help of COMSOL Multiphysics, which is complex modelling software for computational fluid dynamics (CFD). The computational study has been performed with the following parameters, Rayleigh number ( Ra ) = 10 ³ –10 ⁶ , Hartmann number ( Ha ) = 0–80, and nanoparticles volume fraction (ϕ) = 0.01, 0.02. The effect of important variables, such as Hartmann and Rayleigh numbers, in conjunction with the concentration of nano additives has been examined by analyzing streamlines and isotherms to understand their effect on thermal convection. It is found from the isotherms within the cavity in Case 2, that increment in Ha leads to slight rise the temperature within the cavity. Further, in Case 1, Nu avg is decreasing function of Ha and Q . While in Case 2, the average Nu is decreasing function of Q and increasing function of Ra and ϕ.

Emotion and intensity prediction requires a comprehensive understanding of the role emotions play in human communication, as well as advances in natural language processing. Using the EmoInt and EmoBank datasets, this study tackles key challenges in emotion recognition by focusing on emotion classification, intensity prediction, and the analysis of emotional dimensions such as valence, arousal, and dominance. The EmoInt dataset highlights emotion intensity, while EmoBank provides a comprehensive view of emotional variables, making them ideal for advancing emotion prediction research. These datasets were chosen for their thorough coverage of emotional expressions, with EmoInt focusing on emotion intensity and EmoBank providing a rich resource for valence, arousal, and dominance annotations. To achieve these objectives, we provide a new Hybrid Multi-Task BERT_BiLSTM Deep Learning Model (HMT-BB) that predicts emotions and their intensity from textual input. This model uses BERT for contextual comprehension and BiLSTM to represent sequential dependencies. As a result, it excels at capturing subtle nuances of emotional expression. The combination of these two components allows the model to correctly handle the complexities of emotion prediction. The HMT-BB achieves an accuracy of 92.44% in emotion classification on the EmoInt dataset, indicating its capacity to categorize emotions. The model achieves Pearson correlation coefficients of 0.698 for Valence, 0.399 for Arousal, and 0.298 for Dominance on the EmoBank dataset, demonstrating its ability to predict emotional variables accurately. This study proposes a strong and promising approach to emotion and intensity prediction in textual data, bridging the gap between contextual comprehension and sequential dependency modeling. The HMT-BB model has the potential to be a significant tool in applications such as sentiment analysis, affective computing, and natural language processing, where nuanced emotion recognition is critical.

Internet of Things has been increasingly well-known in recent days as a result of its numerous applications. The Internet of Things plays a pivotal role in transforming traditional agricultural practices by enabling smart and precise plant disease detection and classification. Through a network of interconnected sensors and devices, the Internet of Things facilitates real-time monitoring and data collection from agricultural fields. Yet, the traditional detection techniques are limited by poor accuracy, high computational complexity, class imbalance, and overfitting issues. This work introduces a hybrid deep learning methodology known as PlantDetectNet with the IoT to overcome the above difficulties and achieve enhanced outcomes in the plant disease detection and classification process. In this research, the proposed framework employs sensor data and PlantVillage dataset images for accurate disease identification. In the proposed scheme, a gated recurrent unit is applied for extracting the sensor data’s temporal features and Depthcat convolutional neural network is utilized for extracting the spatial features from the input data. The Global Visual Geometry Group 16 framework is employed for mitigating the overfitting, a number of parameters, and refining the intermediate layer features. The Gated ConvNeXt model is utilized for enhancing the classification outcome of the model and effectively gathering the information in the modeling of channel-wise relationships. Additionally, the research introduces a Residual DenseNet approach for eliminating the invalid features and improving the significant features. The experimental results show that the proposed framework attained a high accuracy of 98.8% and a higher recall of 95.9% compared to existing approaches. These simulation findings prove that the proposed methodology enhanced efficiency, accuracy, and scalability in leaf disease detection and improved crop yields.

This paper presents a novel machine learning-driven approach for designing and optimizing multi-band patch antennas tailored for next-generation Internet of Things applications in 5G and 6G wireless communication systems. Recognizing the limitations of traditional antenna design methods, this work leverages the power of ML to enhance antenna performance and design efficiency. We investigate various ML algorithms, including Decision Tree, Random Forest, ANN, KNN, Extra Tree, CatBoost, Gradient Boost, and XGBoost, for predicting antenna characteristics. Notably, the CatBoost algorithm demonstrates superior performance, achieving 77.4% accuracy in predicting antenna return loss. To validate the efficacy of this approach, a multi-band antenna operating across the 3.5–7.8 GHz, 8.5–10.2 GHz, and 11.8–15 GHz frequency bands was fabricated and evaluated. Results demonstrate a good agreement between predicted and measured performance, highlighting the accuracy and efficiency of the ML-driven design methodology. This approach holds significant promise for accelerating the development of high-performance antennas for a wide range of applications, including Wi-Fi, Fixed Wireless Access, Wideband Aeronautical Intranet Communications, IoT devices, Industrial IoT, smart cities, and remote monitoring systems.

The confined constrained deformation behavior of SLM 3D printed SS316 has been analyzed at room temperature, 200 °C, 400 °C, 600 °C, and 800 °C temperature through static ball indentation finite element analysis technique model (load range 5 kN to 50 kN). The constrained confined deformation behavior analysis of SLM 3D printed SS316L steel specimen through static indentation process helps understand their mechanical response under localized compressive loading. This study helps to understand material plastic flow behavior at the time of low-velocity foreign object strike. Further, the results were analysed in terms of Meyer’s hardness (HM), constraint factor (CF), lip height, average strain, strain hardening index (p), and indentation strength coefficient (A). The finite element analysis (FEA) results are validated with the analytical models (Johnson’s Expansion cavity model (ECM) and Richmond’s fully-plastic model (FPM)) and experimental results. The results revealed that the Meyer’s hardness was reduced 10.20%, 21.68%, 42.07% and 66.62% at 200 °C, 400 °C, 600 °C, and 800 °C respectively, compared with room temperature. Further, the lip height was increased by 24.34%, 51.40%, 98.29% and 129.22% at 200 °C, 400 °C, 600 °C, and 800 °C respectively, compared with room temperature. The constraint factor (CF) was 2.539, 2.625, 2.711, 2.961 and 3.211 at room temperature, 200 °C, 400 °C, 600 °C, and 800 °C temperature respectively under confined constrained deformation condition. Further, the CF was increased with increase in the temperature. The yield strength of the investigated material was decreased by 8.60%, 27.08%, 55.44% and 75.45% at the 200 °C, 400 °C, 600 °C, and 800 °C respectively. Further, the compressive strength was decreased by 9.94%, 30.01%, 61.72% and 88.41% at the 200 °C, 400 °C, 600 °C and 800 °C respectively. The FEA model results show the good agreement with the experimental results with less than 5% percentage difference. This shows the good prediction capability of the developed FEA model. The constrained confined deformation behavior study on SLM 3D printed SS316L study is the prime focus of the current investigation which confirms the suitability of material under foreign objects impact conditions like aerospace and defence sector.

Diseases are one of the major factors that have the potential to reduce plant production, food security and ultimately humanity's survival. Therefore, timely and correct identification of plant diseases is important in ascertaining methods to control the diseases. This paper focuses on the application of Deep Learning in identifying plant diseases, and the research's recommendation is a combination of the Vision Transformer (ViT) and GoogLeNet architectures. The objective of this work is to combine the strengths of both models so as to attain increased accuracy and faster computation. This proves that the proposed model yields a substantial accuracy of 99.20% a, 99.30% precision and 99.10% recall. F1‐score shows the highest performance compared to several state‐of‐the‐art models. For comparison, the Vision Transformer, better known as ViT, attained a 97.80% accuracy, 97.90% precision, 97.70% recall and 97.80% F1 scores, and GoogLeNet attained 98. 60% accuracy, 98. 70% precision, 98.50% recall and 98.60% F1‐score. The present hybrid model substantially enhances the capacity to identify plant diseases, hence providing a comprehensive means of managing the early diseases in the plantations. Due to high performance in the desired indicators, it is applicable for real‐world purposes, controlling crops and increasing their yields.