Vidyalankar Institute of Technology

A paradigm shift is occurring in the manufacturing sector globally due to the digitization of operations; however, studies specific to the maintenance aspect are scarce. The research presents challenges particular to the implementation of smart maintenance in Indian manufacturing SMEs, which is unique and helps bridge the gap. This research aims to identify and analyze the hierarchical interrelationship among critical barriers to smart maintenance implementation in Indian manufacturing Small to Medium-sized Enterprises (SMEs). This study involves a two-phase approach. In the first phase, the critical barriers of smart maintenance in Indian manufacturing SMEs are identified. A literature review, along with the opinions of Industrial and academic experts, resulted in the finalization of the barriers. In the second phase, the Interpretive Structural Modeling (ISM) and Cross-Impact Matrix Multiplication Applied to Classification (MICMAC) methodologies were employed to analyze the criticality of the barriers. The driving dependence power was computed for each factor involved, which was then given as an input into the MICMAC analysis to cluster the barriers into autonomous, dependent, linkage, and independent barriers. Smart maintenance is gaining traction worldwide in the manufacturing sector, driven by the emergence and popularity of Industry 4.0. The study identifies the critical hindrances to the adoption of innovative practices specific to the maintenance function of Indian SMEs. ‘Limited Interoperability of Digital Systems’ and ‘Poor Data Quality and Fragmented Data Infrastructure’ emerged as the critical barriers with high driving power. At the same time, ‘Management Resistance and Low Strategic Priority’ and ‘Lack of Cross-functional Collaboration’ appear at the top of the ISM model with high dependence power. The overall approach of smart maintenance, as presented in this research, challenges identification and structured modelling, and is expected to help managers and policymakers concentrate their efforts based on the criticality of barriers. In turn, this will facilitate a smooth transition to smart technologies in the maintenance function of Indian manufacturing SMEs.

Crop yield forecasting is very important for agricultural planning and resource management. This paper presents an innovative approach called R2U-Net-AgriFocus to improve crop productivity forecasting by integrating satellite images and agro-environmental data through a hybrid deep learning architecture. The methodology consists of four main steps. Data pre-processing includes atmospheric correction of satellite images, Histogram Equalization (HE) techniques, and a rotational focus-based recurrent residual U-Net (R2 U-Net). Farm count data were cleaned, outliers were removed, and missing values were imputed by mean imputation. Feature extraction uses a pre-trained Convolutional Neural Network (CNN) model (VGG16) to extract spatial features from satellite images and obtain crop health indices and texture patterns using Gabor filters. Field-specific characteristics are derived from digital data, including historical production trends, soil nutrient levels, and weather data. The Hybrid Hippo-Pufferfish Algorithm (HHPA) is proposed for feature selection to improve model performance. Finally, a DL architecture, R2U-Net-AgriFocus, combines a modified CNN with Convolutional Self-Attention (CSA) inputs and dense blocks, followed by a Gated Recurrent Unit with Recurrent Neural Network (GRUResUnit). CSA models dynamically learn the importance of image features, while dense blocks improve feature reuse. GRU-ResUnit detects long-term dependencies in temporal agricultural data, providing faster training times than traditional Long Short-Term Memory LSTM networks. This method shows significant results in crop yield prediction using the complementary power of agricultural and recurrent neural networks, providing a comprehensive framework for agricultural decision support systems. As such, future work will explore other variations of the proposed model that will be lightweight and suitable for use in resource-constrained agricultural settings in the real world.

This paper presents a comprehensive comparative study of five state-of-the-art CNN architectures, VGG19, ResNet50, InceptionV3, DenseNet121, and EfficientNetB0 for multi-class classification of Chest X-ray images (CXR) into four categories: Edema, Normal, Pneumonia, and Tuberculosis (TB). The models were trained, validated, and tested on a dataset comprising 6,092 training and 325 testing images across four distinct classes. Each architecture was initialized with ImageNet weights, augmented with a custom classifier, and fine-tuned under identical conditions to ensure a fair comparison. The models are evaluated on a comprehensive set of metrics, including accuracy, per-class recall, training time, and model complexity. Experimental results indicate that VGG19 achieved the highest classification accuracy of 98.15%, followed closely by ResNet50 at 97.54%. This study provides empirical evidence to guide the selection of appropriate deep learning models for chest X-ray diagnosis, balancing performance with operational constraints

Cyberbullying is defined as deliberate behavior by an individual or group using electronic means of communication, such as transmitting a message or publishing criticism about a victim on social media in order to harm or slander the victim. Cyberbullying via social media may occur at any time and from any location, as opposed to conventional bullying, which often happens during face-to-face conversation. Currently teenagers have their identity from their social media accounts and are fluently affected if any negative response is seen on their accounts. This research aims to mitigate cyberbullying. They often feel discouraged, disheartened, and hurt when subjected to aggressive responses to their posts or harsh, targeted communications. In this work, we’ve probed various algorithms to describe cyberbullying and compare them to get the best result. Experimenters have extensively used SVM and CNN for the discovery of cyberbullying on various social platforms. We explore the use of text feature extraction methods—Count Vectorizer (CV) and Term Frequency-Inverse Document Frequency (TF-IDF)—combined with various machine learning algorithms for real-time detection of cyberbullying. By assigning importance to words beyond their raw frequency, TF-IDF demonstrated superior performance across most algorithms compared to CV. Algorithms like Linear SVC and Random Forest performed exceptionally well. This approach can be instrumental in developing efficient, real-time cyberbullying detection systems, particularly on social media platforms.

In the intricate landscape of dermatological diagnostics, traditional methodologies often necessitate the specialized expertise of healthcare professionals to discern between subtly varying skin conditions. To bridge this gap, our research advances a computational framework that amalgamates classical convolutional neural networks (CNNs) with custom-designed layers for nuanced skin lesion classification. The model accommodates 224 × 224 pixel images across three color channels (RGB) and counters inconsistencies such as variable lighting through data augmentation techniques like random contrast adjustments. Established architectures like VGG16 and ResNet101 are employed via transfer learning for initial feature extraction. Unique to our model are custom layers: a Fourier layer for frequency domain characteristics, a Laplacian layer for edge detection, a Gabor layer for texture analysis, an RGB-to-HSV layer for color variation, and a histogram equalization layer for image contrast. These are aggregated in a ‘CombinedOutput’ layer and processed by feed-forward layer-X and layer-A using GELU activation and batch normalization. A final concatenation layer leads to a softmax-activated prediction layer. We employed exhaustive performance evaluation metrics: Accuracy, Precision, Recall, AUC, F1. Our best-performing custom model exhibited a Recall of 94.2