Grinding is a critical machining process used to achieve superior dimensional accuracy and surface integrity in precision components across the automotive, aerospace, and manufacturing industries. This study introduces a hybrid framework that integrates Taguchi-based empirical modeling with Non-dominated Sorting Genetic Algorithm II (NSGA-II) optimization to assess and improve grinding performance while considering dressing parameters within the limitations of practical experimental settings. The parameters studied were dresser depth of cut, dresser cross-feed rate, and grinding feed rate. Experiments were designed using a Taguchi L9 orthogonal array with two replicates per run, yielding 18 trials in total. Surface roughness, grinding power, and grinding ratio were selected as performance objectives. By focusing on the simultaneous optimization of surface roughness, power consumption, and grinding ratio, this framework distinguishes itself by including dressing effects in a multi-objective evolutionary model. The hybrid framework successfully generated Pareto-optimal solutions for conflicting objectives. The best result achieved a surface roughness of 0.2795 μm, with 1.696 kW grinding power and a grinding ratio of 11.1621. The use of a resource-efficient experimental design further enhances its practical applicability, offering sustainable and effective solutions for industrial-grinding processes.
Fusion-based welding techniques for joining aluminium or its alloys are widely used in industries. However, their effectiveness is restricted due to the high thermal conductivity, low melting temperature, and strong affinity for oxygen of these materials. These inherent characteristics lead to solidification cracking, porosity, oxide entrapment, elemental segregation, excessive heat-affected zone softening, precipitate coarsening, and high residual tensile stresses, thereby degrading joint integrity and mechanical performance, particularly in heat-treatable alloys. To mitigate these issues, welding parameters optimization and post-weld treatment are employed. In the context of post weld treatment, heat treatment and thermomechanical processing have been in the focus, in which friction stir processing (FSP) has drawn an immense attention of researchers in recent years because of its solid-state and thermo-mechanical characteristics. FSP facilitates severe plastic deformation, dynamic recrystallization, grain refinement, homogeneous microstructure evolution and homogeneous phase distribution in the materials. Hence, when applied to fusion-welded aluminium joints, FSP effectively mitigates solidification-related defects since there is no melting, refines coarse dendritic microstructures due to severe plastic deformation into fine equiaxed grains, redistributes strengthening precipitates, reduces residual stresses, and narrows the effective heat-affected zone. This review paper deals with evaluating the role of FSP as a post-weld treatment technique, emphasizing microstructural evolution and its influence on mechanical and corrosion performance of fusion welds of aluminium alloy.
Rising population impacts the excessive consumption of petroleum fuels leading to global warming and changes in the environment. As a result, it is important to use renewable and clean energy such as biodiesel as a source of energy to fulfil the energy requirement. It leads to complete combustion of fuel and reduces exhaust gas emissions. This study focuses on the green production of low-cost catalysts using discarded fish scale Labeo Rohita and optimised biodiesel production using design expert software. Analytical tools such as Xray diffraction (XRD), Fourier Transform Infrared spectroscopy (FTIR), and Scanning Electron Microscopy (SEM) are used to analyse synthesized catalysts. At optimal conditions, 90.21% yield of biodiesel was obtained at 1.5 wt.% catalyst concentration, at temperature of 65 0 C, and 1.5-hour reaction time. Several properties of prepared biodiesel such as density, flash point, pour points, and kinetic energy were evaluated and validated as per American Society for Testing and Materials (ASTM) Standards.
The conventional method of visually inspecting tea crops is time-consuming, requires human judgement, and is unsuitable for large scale tea plantations. This paper proposes a deep learning automated framework to detect tea leaf diseases using Convolutional Neural Networks (CNN) combined with Generative Adversarial Networks (GAN). The GAN will provide realistic synthetic images to augment the training set; this is an effective approach to addressing class imbalance and improve the generalization of the model. The study finds that the CNN captures the discriminatory spatial characteristics of the tea leaf images and classifies the leaf images into three categories of diseases. The framework’s classification results indicate an overall accuracy of 97.6%, precision of 96.9%, recall of 97.8%, F1 score of 97.3%, and specificity of 97.1%. The study also finds that the proposed CNN-GAN framework greatly improves the robustness and classification accuracy of tea disease detection compared to traditional CNN, ResNet-50, Inception-v3 and EfficientNet-B3. Additionally, the proposed CNN-GAN framework provides efficient inference time which can be employed in real-life applications and highlight GAN assisted deep learning methods for achieving accurate results at low cost.
Geological modelling plays a critical role by improving subsurface characterization and supporting engineering decision-making. This study presents an integrated geostatistical modelling framework for spatial characterization of lithology, core recovery, rock quality designation (RQD) and percolation geoparameters at Bhama Askhed Irrigation Project site, Pune, India. Unlike conventional applications of geostatistical interpolation, the present work combines three-dimensional geological modelling, statistical validation, engineering performance assessment, and post-failure verification using Energy Dissipating Assembly (EDA) failure investigations by the Central Water and Power Research Station (CWPRS), Pune. Model performance was evaluated using correlation coefficient (R), Root Mean Square Error (RMSE), Nash-Sutcliffe Efficiency (NSE), and Ratio of Root Mean Square Error to Standard Deviation (RSR). The developed models exhibited satisfactory agreement between observed and predicted data with low residual errors and acceptable performance indices. Solid model residuals for the interpolated model are substantially less: 6.15, 8.29, 13.07 and -9.60 for lithology, recovery, RQD and percolation, respectively, thus indicating reasonable model fit. Statistical comparison of model-predicted parameters with the reserved data set also shows correlation coefficient greater than +0.70 indicates a strong uphill linear relationship between the observed and modelled data. The average correlation for lithology, core recovery, RQD and percolation was found to be 62.25%, 92.5%, 88.25% and 75.5%, respectively. The conducted tail channel erosion modelling can identify intensity of erosion-induced geological deterioration, enabling early recognition of vulnerable zones. The predicted weak zones showed consistency with field observations and post-failure investigations, highlighting the practical engineering applicability of the proposed framework. The study demonstrates that integrated geostatistical-geological modelling can, improve spatial prediction of geological parameters, produce missing spatial investigation data, support engineering design and risk assessment in infrastructure projects.
