Deep learning–based welding defect classification often faces challenges such as limited training data, class imbalance, and high model complexity, which restrict real-time industrial applications. To address these issues, this paper proposes a hybrid CNN–Vision Transformer framework with GAN-based data augmentation for welding defect classification. First, welding images acquired using a wide dynamic range visual sensor are preprocessed through binarization, median filtering, morphological dilation, and cropping to enhance defect features. A Generative Adversarial Network (GAN) is then employed to generate synthetic samples and alleviate dataset imbalance. A lightweight CNN extracts local spatial features and reduces feature dimensionality, after which the resulting feature maps are converted into tokens and processed by a Vision Transformer encoder to capture global contextual relationships via self-attention. The proposed model classifies welding images into four categories: normal, burn-through, undercut, and welding collapse. Experimental results demonstrate that the hybrid architecture achieves improved classification accuracy and computational efficiency compared with conventional lightweight CNN models. In addition, the model attains 98.25% accuracy on the MNIST dataset, validating the effectiveness of the proposed framework.

The article examines the comparative properties of Retrieval-Augmented Generation algorithms and large-language-model fine-tuning methods in the context of domain-specific search tasks with a high cost of error. The aim is to identify operating regimes in which RAG and fine-tuning differentially affect the accuracy of top-ranked results, the evidential quality of answers, and the safety of handling sensitive data. The relevance of the study is driven by the rapid growth of industrial domain-specific search systems that must simultaneously ensure knowledge updatability, strict citation-based verifiability, and regulatory discipline. The novelty lies in the fact that the comparison is conducted not in the abstract form of RAG versus fine-tuning, but at the level of individual pipeline components and from the perspective of operational trade-offs: it is shown that retrieval and ranking form a truth scaffold and a channel for knowledge refresh, whereas fine-tuning acts as a delicate regulator of format, terminology, and epistemic precision without resolving the problem of obsolescence in parametric representations. The article concludes in favor of hybrid schemes that combine hybrid retrieval, reranking, and strict citation rules with lightweight, parameter-efficient model adaptations, thereby enabling reproducible, controllable, and scalable operation of domain-specific search systems. The article is intended for researchers in information retrieval, engineers of applied RAG systems, and practitioners deploying generative models in high-risk domains.

Industry 4.0 has increased the demand for intelligent predictive maintenance systems capable of supporting real-time monitoring, early fault detection, and efficient decision-making in industrial environments. In this context, accurate prediction of machine failures has become essential for minimizing downtime, reducing maintenance costs, and improving operational reliability. This study employs the Predictive Maintenance Dataset from the UCI repository to develop and evaluate data-driven models for machine failure prediction and classification. The research pursues two primary objectives: first, to compare the performance of several machine learning algorithms in classifying machine failures, and second, to assess the effectiveness of deep learning approaches in achieving higher predictive accuracy. Among the machine learning models examined, the XGBoost classifier demonstrates the strongest performance. To further enhance prediction capability, this study adopts a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) model, which integrates CNN’s strength in automatic feature extraction with LSTM’s ability to learn temporal dependencies from sequential data. Experimental results show that the proposed CNN–LSTM model outperforms traditional machine learning models as well as Artificial Neural Networks (ANN) in predicting machine failures. The main contribution of this study lies in the comparative evaluation of machine learning and hybrid deep learning techniques on an imbalanced predictive maintenance dataset. The findings confirm the potential of hybrid deep learning models for predictive maintenance applications and highlight their practical value in enabling proactive maintenance strategies, optimizing resource allocation, and enhancing the reliability of smart industrial systems.

Modern healthcare, research, and data center facilities require highly reliable HVAC systems capable of maintaining strict environmental conditions and pressure relationships. Inadequate airflow distribution, hydronic imbalance, and poor pressure control can compromise system performance and safety in mission-critical environments. Testing, Adjusting, and Balancing (TAB) provides a structured methodology to validate and optimize system performance.

This paper presents an integrated TAB framework that includes air distribution balancing, chilled water hydronic balancing, liquid cooling system balancing, and pressurization verification for critical rooms. Field measurements were evaluated using descriptive statistics and regression analysis to quantify performance improvements.

Results indicate airflow deviation reductions of up to 74%, chilled water flow variation reductions of 72%, liquid cooling efficiency improvements of approximately 12%, and pressure stability improvements within ±0.01 in. w.g. A case study conducted at Sidra Medicine in Doha demonstrates the application of this methodology in a large healthcare and research facility. The findings highlight TAB as an essential component of commissioning and performance validation in complex HVAC systems. 

This paper presents an inspection-based civil case study of drainage vulnerability during active construction. Using two civil field inspection reports, a coordination log, response emails, and supporting site photographs from a mid-construction residential project in Florida, the study examines how drainage-related deficiencies developed after stormwater infrastructure was already materially installed, how those deficiencies were addressed, and what those events reveal about preserving drainage function before final site stabilization. The record shows a progression from control slippage to direct system vulnerability, including temporary exposure of active drainage elements, sediment intrusion into collection features, uncertainty in an internal manhole control detail, breakdown of work-area sediment containment, and localized loss of drainage effectiveness caused by sediment accumulation. These conditions are interpreted as linked preservation failures within a live drainage network rather than as isolated field defects. The response record further shows that different conditions required different closure pathways, including immediate field protection, correction and reconciliation of a questioned internal detail, and reinforcement of site-scale sediment controls. On that basis, the paper distinguishes between observed deficiency, documented corrective action, and construction-control closure. The study concludes that the principal construction-stage risk lies not in incomplete installation alone, but in loss of protection around drainage assets that have already become operational, and it proposes a practical framework for preserving those assets until the site reaches stable completion.