Automated visual inspection plays a central role in electronics manufacturing processes involving UV adhesive deposition, where undetected defects may compromise mechanical stability and lead to latent failures. In NVMe solid-state drive (SSD) production, inspection systems must prioritize reliability and defect containment, as false negatives represent a critical operational risk. Despite this requirement, many deep learning-based inspection approaches remain optimized for global accuracy metrics, which are not fully aligned with industrial reliability constraints. This paper presents a convolutional neural network (CNN)-based visual inspection framework tailored for high-reliability deployment in NVMe SSD manufacturing. The proposed approach emphasizes inspection problem formulation rather than architectural complexity. A physically coherent region of interest (ROI) was defined to encompass the functional UV adhesive deposition area surrounding the SSD controller, reducing background interference. Three CNN backbones—ResNet50, EfficientNetV2, and MobileNetV2—were evaluated under identical conditions using transfer learning. Additionally, multiple decision strategies, including calibrated decision thresholds, were analyzed to reduce false negatives while controlling false-positive rates. All experiments were conducted using an industrial dataset collected from an operational production line. Experimental results indicate that ResNet50 achieved stable accuracy at around 86% but showed limited reliability due to elevated false-positive rates. In contrast, EfficientNetV2 and MobileNetV2 achieved a defective part recall above 98%, overall accuracy exceeding 91%, and a reduction of more than 50% in false negatives compared to the ResNet50 baseline. MobileNetV2 matched the performance of EfficientNetV2 while maintaining lower computational complexity. The results demonstrate that reliability gains in industrial visual inspection are more effectively achieved through region-focused analysis and decision-level optimization than by increasing model complexity. The proposed framework is suitable for deployment in high-throughput manufacturing environments and provides a basis for future extensions involving hybrid inspection strategies.

Finding the complete set of stabilizing parameters for a Proportional–Integral–Derivative (PID) controller is a long-standing challenge in control engineering, especially when dealing with high-order systems. This study proposes a practical computational framework to map the entire stability region in the (Kp,Ki,Kd) parameter space utilizing the singular frequency decoupling method. By fixing the proportional gain (Kp), the boundaries of the stability region in the (Ki,Kd) plane are analytically derived using linear equations. This approach allows for the automatic identification of stabilizing polygons without the need for complex and time-consuming grid-based searches. A key focus of this work is the extension of this methodology to handle parametric uncertainties within interval plants. Instead of relying solely on traditional methods, this study employs a robust analysis based on vertex plant configurations. "By evaluating the stabilizing proportional gain intervals for the extreme corners of the uncertainty box and finding their mathematical intersection, a common solution region is identified. Furthermore, robust stability within the (Ki, Kd) plane is ensured by intersecting the stabilising polygons of the individual vertex plants and conducting a robust stability analysis of the parameter space, taking into account the uncertain plant parameters." To validate the approach, various representative system models are examined and analyzed. The results demonstrate that the proposed method provides a reliable and efficient tool for designers to determine robust controller gains with guaranteed stability, visualized through 2D robust polygons and 3D stabilizing solids.

Industry 4.0 drives the evolution toward efficient, intelligent, and interconnected production systems, where standardized digital twins—centered on the Asset Administration Shell (AAS)—provide a unified digital representation of physical and logical assets.

This paper demonstrates a comprehensive digitalization methodology by transforming a legacy industrial electric screwdriver into a fully compliant Industry 4.0 component, serving as a concrete case study. The approach rigorously follows the RAMI 4.0 reference architectural model and the Acatech Industry 4.0 Maturity Index principles, enabling progressive maturity advancement in brownfield environments.

The generic, replicable process consists of six modular stages adaptable to virtually any industrial asset: (i) asset characterization, functional analysis, and digitalization objective definition; (ii) creation of a Type 1 AAS (static/digital master) using standardized submodel templates for semantic description; (iii) design and deployment of low-cost/custom IoT sensing hardware to capture relevant real-time data (e.g., energy, usage, condition); (iv) bidirectional integration linking the physical asset to its digital representation; (v) implementation of a dynamic Type 2 AAS with secure runtime interfaces (e.g., OPC UA server); and (vi) real-time data access, visualization, and analytics via standardized clients.

This standardized, scalable methodology offers a practical blueprint for retrofitting legacy equipment without requiring full system replacement, thereby accelerating Industry 4.0 adoption across diverse manufacturing domains. The screwdriver implementation validates how standardized digital twins enable enhanced condition monitoring, energy transparency, predictive insights, data-driven decision-making, and improved operational efficiency and sustainability.

Underground mining environments pose significant safety risks due to the accumulation of toxic and explosive gases, necessitating the use of reliable and continuous monitoring systems. However, wired system configurations, complex maintenance, and incomplete spatial coverage have limited the implication of conventional methods, which hinder real-time data transmission during critical events, thereby increasing risks to miners and delaying emergency response. Herein, a ProGas-Mine System, a low-power, multi-node wireless sensor system, was developed to enable continuous and real-time monitoring of hazardous gases in underground mines, which can communicate over a long range. The proposed system integrates distributed sensor nodes equipped with gas sensors for methane (CH₄), carbon monoxide (CO), and hydrogen sulfide (H₂S), as well as temperature and humidity sensors. The incorporated Long Range (LoRa) wireless communication ensured reliable data transmission under challenging subterranean conditions. DC–DC switching converters and a battery fuel-gauge-based power management system were further developed to enable the system to operate in an energy-efficient manner. The technical functionality and real-time performance of the system were experimentally validated in a large-scale commercial graphite mine, precisely demonstrating accurate, stable, and continuous sensing of gas concentrations and environmental parameters. Moreover, the results revealed that the system exhibits strong responsiveness, capturing real-time variations in gas levels that correlate closely with the mine ventilation cycles. Accordingly, the developed system provides a scalable and cost-effective alternative to traditional monitoring methods, facilitating a transition from reactive safety measures to proactive and predictive risk management in underground mining environments.

The increasing demand for automation in the industry necessitates innovative solutions for efficient material handling. This work introduces the design of a robotic hand gripper for pick-and-place operations.

The design employed a multi-fingered design that integrates compliance with precision actuation mechanisms to enhance object manipulation capabilities. The gripper’s structure utilizes lightweight yet durable materials, optimizing performance while minimizing energy consumption. The methodology includes computer aided design modeling and simulation to analyze the gripper's mechanics, followed by the fabrication of a prototype using 3D printing mechanisms. Incorporated as well force sensors and adaptive control algorithms to ensure real-time feedback and adjust grip strength according to various object characteristics.

Experimental results demonstrated that the robotic gripper can successfully handle objects weighing up to 5 kg with a gripping precision of ±2 mm. The adaptability of the gripper allows it to securely grasp items ranging from cylindrical bottles to irregularly shaped containers, showcasing its versatility in diverse operational scenarios.

In conclusion, the proposed robotic hand gripper significantly enhances the efficiency and reliability of pick-and-place tasks in automated workflows. Future work will focus on refining the control algorithms and expanding the gripper’s capabilities to include more complex tasks, paving the way for its application in a wide range of industrial contexts.

Modern electromechanical systems increasingly operate in dynamic, uncertain environments where conventional control and monitoring pipelines are insufficient for ensuring reliability, efficiency, and autonomy. Most existing approaches treat condition monitoring and control as isolated processes and rely heavily on static models and offline analysis. This paper introduces a unified neuro-adaptive machine framework designed to endow machines with real-time perception, health awareness, and self-optimizing control capabilities directly at the edge.

The proposed architecture integrates multi-modal sensor fusion (vibration, acoustic, thermal, and electrical signals) with a lightweight deep learning pipeline deployed on embedded hardware. A sparse neural representation layer performs continuous feature extraction, while a continual learning module tracks machine state evolution and detects emerging fault patterns. These learned health states are coupled with a predictive control module that dynamically adjusts operating parameters to mitigate degradation. Drift-aware training and incremental updating enable the system to adapt autonomously to changing operating conditions without cloud dependency.

Experimental evaluation on an electromechanical test platform demonstrates that the framework achieves earlier fault detection, higher diagnostic accuracy, and faster adaptation to unseen anomalies compared to traditional signal-processing-based monitoring and fixed-parameter control strategies. The system also exhibits improved operational stability and measurable gains in energy efficiency through health-aware control adjustments.

The results validate the feasibility of embedding neuro-adaptive intelligence directly into machines to move from passive monitoring toward proactive, self-optimizing operation. The proposed framework offers a scalable pathway toward cognitive electromechanical systems, contributing to advances in automation, condition monitoring, and intelligent machine design.

Within the contemporary paradigm of escalating intricacy within waste supply chains, the utilization of simulation as a modeling instrument for existing processes within supply chains has assumed a pivotal role. This instrument is utilized for the analysis and generation of scenarios, with the objective of predicting results and performance indicators in terms of sustainability. Nevertheless, a plethora of studies concentrate on the reduction of costs and collection and treatment times, emphasizing operational and logistical performance, while human behavior is regarded as a constant variable. However, this approach is not without its limitations, particularly when it comes to the applicability of models in real-life settings. In such contexts, it becomes imperative to incorporate societal behaviors, such as adherence to sustainable policies, into the modeling process. It is therefore crucial to make the model dynamic and more realistic. The present article thus aims primarily to examine the societal contribution to the simulation of waste supply chains through a literature review of concepts such as the role of society in sustainability and simulation. Subsequently, a quantitative and qualitative analysis was performed using the Scopus database. The results of the study indicate that a mere 6% of the documents under review are associated with the social dimension of sustainability, a figure that is clearly insufficient given the importance of the social dimension. It is recommended that further studies be conducted in order to apply this dimension of sustainability to simulation, with a view to obtaining more complex and robust models and, consequently, more accurate results.

This paper presents an embedded vision-based autonomous system designed to convert hand-drawn sketches created on a transparent glass surface into engraved or cut patterns on solid materials such as wood and plastic. The proposed machine aims to simplify human–machine interaction in digital fabrication by enabling users to draw naturally by hand without requiring a computer, display, or specialized software interface.

The system integrates a glass-based drawing surface positioned above an embedded camera that captures the user’s sketch from below. The acquired image is processed locally on a Raspberry Pi, where embedded image processing algorithms are applied to extract contours and geometric features from the hand-drawn sketch. The extracted paths are then converted into standard G-code instructions, ensuring compatibility with conventional CNC motion control principles.

The generated G-code is executed directly by a dedicated mechatronic platform consisting of a Cartesian motion system driven by stepper motors. To ensure positional accuracy and repeatability, a reference positioning sensor is employed to define consistent machine origin prior to each operation. After pressing a single physical start button, the system autonomously reproduces the original hand-drawn sketch as an engraved or cut pattern on the target material without further user intervention.

Unlike conventional CNC or laser engraving systems that depend on external computers and complex graphical interfaces, the proposed solution emphasizes autonomy, usability, and compact system integration. Experimental results demonstrate reliable reproduction of complex hand-drawn shapes with stable motion behavior and repeatable positioning accuracy.

The proposed approach is particularly suitable for educational environments, artistic fabrication, and low-cost rapid prototyping applications, and highlights the potential of embedded vision and autonomous mechatronic systems in human-centered manufacturing workflows.

Maritime piracy continues to pose a serious threat to commercial shipping, especially in regions where surveillance coverage is limited and response time is critical. Conventional anti-piracy measures rely primarily on human watchkeeping and isolated monitoring systems, which may be insufficient for early threat recognition. The rapid development of autonomous systems and artificial intelligence enables new approaches to maritime situational awareness based on continuous, automated analysis of multisource data.

Here, we present a conceptual design of an autonomous early-warning system for maritime piracy threat detection based on AI-driven sensor fusion. The proposed framework integrates heterogeneous data from Automatic Identification System (AIS), marine radar, and electro-optical/infrared (EO/IR) sensors into a unified operational picture. A data fusion layer is combined with machine learning algorithms for anomaly detection and vessel behavior classification, enabling identification of potentially hostile units and atypical navigation patterns. The system architecture includes an onboard edge-computing unit responsible for real-time data processing and automatic generation of threat alerts for the ship’s bridge team.

The article discusses key design requirements, data processing pipelines, decision-making logic, and integration with existing shipboard control and communication systems. Particular attention is paid to autonomy, reliability, and reducing false-alarm rates in complex maritime environments. The proposed concept demonstrates how AI-based sensor fusion can enhance early threat awareness and support crew decision-making in piracy-prone waters.

The presented framework provides a foundation for future development and practical implementation of autonomous maritime security systems on commercial vessels.

Supervisory Control and Data Acquisition (SCADA) systems play a key role in various industrial processes. Due to their key role in critical infrastructures, in recent years, they have become the target of cyber attackers. Although some security hardening solutions have been proposed to secure them, traditional security measures such as firewalls, intrusion detection systems, and access controls are not enough to provide adequate protection against modern cyber threats. Therefore, there is a need for novel security hardening solutions that can detect and respond to emerging, previously unknown threats. In parallel with this, in this research, we propose an intelligent security hardening approach for SCADA systems using machine learning algorithms. The proposed approach relies on the collection and analysis of network traffic data from SCADA systems, followed by the application of machine learning algorithms to detect and respond to cyber threats. Network traffic data collected from various sources are analyzed to identify anomalies that may indicate the presence of cyber threats. Various machine learning algorithms are used to analyze the data. The proposed approach can improve the security of SCADA systems and reduce the risk of downtime and financial losses due to cyber attacks. It is a more cost-effective security solution compared to traditional security measures.