Development of an automated fruit classification system by using computer vision and deep learning.

Towards responsible innovation ecosystems through social labs: Lessons from the field

Recent developments in software engineering for systems-of-systems and software ecosystems

Grasp strategy-driven design of soft robotic grippers for food industry applications

AutoGraph: An intelligent knowledge-graph agent for procedure automation and dynamic human reliability support in high-risk industries

The rapid proliferation of deep learning models across critical socio-technical infrastructures has necessitated a paradigm shift from purely performance-oriented development to security-centric architectural design. As deep learning systems transition from controlled laboratory settings to real-time, high-stakes environments—such as autonomous transportation, industrial automation, and financial grid management—they become increasingly vulnerable to adaptive adversarial attacks [13, 19]. These attacks exploit the inherent brittleness of high-dimensional neural networks through strategically crafted perturbations designed to deceive model logic while remaining imperceptible to traditional monitoring systems [5, 11]. This research paper explores the systemic integration of certified robustness training as a foundational security layer for deep learning infrastructures. Unlike empirical defenses that rely on heuristic methods and often fail against novel or adaptive threats, certified robustness provides a mathematically grounded guarantee of model stability within defined perturbation bounds [8, 14]. The study analyzes the structural trade-offs between computational overhead, certified accuracy, and real-time latency requirements. Furthermore, it examines the governance and policy implications of deploying such robust systems, emphasizing the need for standardized certification protocols in public infrastructure. By synthesizing insights from systems engineering, cybersecurity, and algorithmic fairness, this paper proposes a holistic framework for resilient artificial intelligence deployment, ensuring that the next generation of automated systems remains reliable and secure against the evolving landscape of sophisticated adversarial interference.

Miniaturization of high-performance antennas for sub-6GHz applications such as 5G small cell, Internet of Things (IoT) devices, wearable medical system and industrial automation is essential in the near future, and the novel design methodology to realize the antenna miniaturization with high efficiency is required. In this context, this paper proposes a nature inspired antenna design using sneezewort leaf shape geometry combined with DGS (defective ground structure) to improve bandwidth and radiation features. A partial ground, the proposed antenna is very compact in size 18 × 19 × 1.6 mm3 (0.18λ × 0.2λ × 0.016λ) and has a wideband operation, that is, 3.16-5.42GHz. It shows wide impedance bandwidth of 55.96% and has a highest gain of 2.1 dBi. Due to its planar, lightweight and low-profile structure, the antenna is ideal for low-cost mass production and easy integration with emerging wireless and healthcare gadgets. The multifunctional antenna lends itself to varied uses such as sub-6GHz 5G communication, high-speed Wi-Fi, near-field vehicular radar, and industrial ISM-band devices. Besides, the antenna clearly show good prospect for many biomedical diagnostic applications, such as breast, brain, skin, lung, heart and kidney abnormalities detection, temporomandibular joint (TMJ) disorders, typhoid, bone fracture, dengue, food contaminations and lately even COVID-19. A breast tumor detection proof-of-concept is demonstrated exclusively by simulation using a breast phantom, which demonstrates the antenna sensitivity to changes in dielectric properties. These results indicate that the proposed antenna could be potential for the wireless communication in the future and also in the medical applications. Experimental validation including studies on physical phantoms and clinical studies will be included in future work.

The research is devoted to solving the urgent problem of industrial production intellectualization based on the creation of an intelligent HMI display using a unified artificial immune system (UAIS) with neuro-endocrine interaction technologies for the control of complex objects and equipment diagnostics in the oil and gas industry. The developed intelligent adaptive display with a predictive alarm system based on a digital twin of the technological process allows reducing the internal and external load on the operator during the operation of high-tech equipment at oil and gas processing plants. Control of the alarm system is of key importance for ensuring safety and maintaining the efficient operation of the production process, i.e. homeostasis. Of extreme importance is the problem of adequate reaction to decision-making by the operator in case of possible failures in the technological process and equipment operation. This issue is especially acute in the automation of large-scale complex production. The control of the alarm system is dynamic and can change depending on many factors. To process multidimensional information about the state of a complex object and predict the behavior of the system, a unified artificial immune system is used in interaction with an artificial neural network to identify informative features when working with historical data, as well as endocrine regulation of homeostasis in the system. With the help of an intelligent HMI display built on these principles, the internal load on the operator is reduced, which allows for effective decision-making on process management. The results of experiments and modeling using the proposed technology and the developed intelligent HMI display on Honeywell Experion PKS equipment in the Honeywell laboratory of the School of Information Technology and Engineering, Kazakh-British Technical University in Almaty, Republic of Kazakhstan are presented. This UAIS technology can be used with industrial equipment from other vendors for SCADA (Supervisory Control and Data Acquisition) and DCS (Distributed Control System) systems with support for web technologies HTML (HyperText Markup Language) and CSS (Cascading Style Sheets) in the development of displays for workstations and operator panels.

The growth of graphene on Cu via chemical vapor deposition has been well established for producing large-area high-quality graphene films, with subsequent transfer to target substrates being an essential step for most applications. While various transfer techniques have been explored, true industrial adoption remains limited. This is primarily because the strong adhesion between graphene and its growth substrate makes it difficult to maintain film integrity by mechanical peeling methods, though they are compatible with industrial operation. Conversely, traditional chemical etching methods, while preserving integrity, rely on highly flexible carrier films to ensure conformal contact with the target substrate. This required flexibility renders the films fragile and difficult to handle, posing a significant challenge for automation and spatial alignment. Here, we report a mesh-embedded Vaseline structure, inspired by reinforced concrete, as a robust carrier film for graphene transfer. The Vaseline acts as an adhesive layer, ensuring conformal contact to preserve graphene's integrity, while also being easily removable. The embedded mesh provides a self-supporting framework, enabling straightforward handling and compatibility with industrial automation. We demonstrate successful graphene transfer onto SiO2/Si wafers and curved surfaces with excellent integrity and cleanliness, and present a semiautomated transfer production line. Leveraging the precise spatial control afforded by the self-supporting carrier film, we further achieve, for the first time in a wet-transfer process, the fabrication of bilayer graphene with precisely controlled twist angles using CVD-grown domains, a capability previously accessible only through small-area manual assembly techniques. This work highlights the potential for scalable mass production and extends readily to other two-dimensional materials such as hexagonal boron nitride. By transforming the transfer process from handling fragile, floating films to manipulating a robust, self-supporting composite, our method bridges the gap between lab-scale demonstration and industry-scale automation.

Computer Numerical Control (CNC) inserts are critical components of CNC machine tools, where surface defects can severely compromise machining precision. Traditional manual inspection methods for these defects are inefficient and prone to significant oversight. To address these limitations, this paper presents an automated real-time system for detecting surface defects on inserts. A dedicated dataset of CNC tool inserts was created and annotated with defect categories. We propose an Attention-Augmented Multi-Defect YOLO model (A2MD-YOLO) for surface defect detection on CNC inserts. The development of this model is motivated by key characteristics of the dataset, which include substantial variation in defect sizes, high intra-class appearance variance, and low inter-class variance. A2MD-YOLO achieves higher detection efficiency and accuracy while reducing the rate of missed detections. The A2MD-YOLO model demonstrates a substantial performance improvement, with the [Formula: see text] increasing from 0.529 to 0.571 and the missed detection rate decreasing from 21.4% to 11.8%. Finally, the proposed algorithm was implemented into the hardware system, enabling automated detection of surface defects on CNC inserts.

This study presents a novel hybrid cryptographic model designed to enhance privacy preservation and data integrity in IoT-enabled Wireless Sensor Networks (WSNs). Traditional algorithms such as RSA, AES, and Blowfish are evaluated and combined into a Hybrid Model to address the resource-constrained nature of IoT devices. The proposed model was tested on a dataset of sensor data, with performance metrics including encryption/decryption time, security strength, memory usage, data throughput, and communication overhead. Numerical findings demonstrate the Hybrid Model's superior performance, with encryption time reduced by 18% compared to Advanced Encryption Standard (AES), The hybrid model employs RSA-2048 (112-bit security strength) for key exchange and AES-256/Blowfish for data encryption (256-bit confidentiality protection). The memory usage was optimized, requiring only 25.16KB, making it suitable for low-power IoT devices. Additionally, the Hybrid Model achieved a data throughput of 24.89KB/s and reduced communication overhead to 1.32KB. These results highlight the efficiency and robustness of the Hybrid Model in securing IoT-enabled WSNs. This research contributes a scalable, resource-efficient solution for privacy and data integrity, offering a promising advancement for real-time IoT applications in sectors such as healthcare, industrial automation, and smart homes.

ABSTRACT - The quality control system perform quality checks of the manufactured products in most of the industries post production using quality control department. The manual inspection of the quality is the error prone task and requires repetitive quality inspection with same precision. This project deals with the development of the visual feature based recognition and segregation for industrial intelligence using Robotics. The proposed project consist of development of vision based system which can check for defects. The products to be inspected for quality and dimension are passed below the camera based quality inspection system where they are check for the quality or dimensional accuracy. The deep learning based system is used to determine the defects and if the product is found to be defective the segregation system will automatically segregate the defective parts from the flow. The proposed system uses deep learning for visual feature inspection for detection of industrial defects as well as dimensional errors in manufacturing. Keywords: Machine Vision, Defect detection,Quality assurance,Industrial automation, Quality inspection system.

This study addresses the inherent limitations of conventional PI control in speed regulation systems, particularly its poor robustness. A dual‐loop model predictive control (MPC) strategy integrating speed and current loops is proposed, treating the motor as a multi‐input multi‐output system to effectively mitigate issues such as weak robustness and slow dynamic response during operation. The speed loop employs Model Predictive Speed Control (MPSC) for speed regulation. However, since MPC performance partially relies on the accuracy of the motor's mathematical model, load disturbances may degrade control performance. To address this, a load observer is introduced for the LIM speed loop to estimate and compensate for disturbances, significantly improving control accuracy. Meanwhile, the current loop employs model predictive flux control (MPFC), enhanced by an improved generalized dual‐vector control strategy to optimize steady‐state performance, thereby effectively reducing current ripple and torque pulsation. Simulation results demonstrate that the proposed dual‐loop MPC strategy exhibits superior performance under speed step changes and load disturbances, featuring minimal overshoot, fast dynamic response, high steady‐state accuracy, and strong anti‐interference capability. Additionally, the observer effectively tracks the given load. This study provides a practical control solution for high‐performance LIM applications in rail transit and industrial automation. © 2026 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.

ABSTRACT In today’s dynamic industrial environment, shaped by volatile markets and resource competition, manufacturers must adapt swiftly to remain viable. These circumstances often demand physical changes that may affect the underlying Control Program (CP). Adapting the CP entails reconfiguring systems between operational states, where the associated complexity and effort are crucial factor for evaluating the economic feasibility of alternative strategies. While effort estimation has been extensively explored in software engineering, existing approaches do not quantify the adaptation effort of CP, particularly in Cyber-Physical Production Systems (CPPSs). This paper introduces a complexity index-based graph model for CP adaptation effort estimation and operationalises it via scaling index-to-time-factor to produce quantitative time and cost estimates under scarce industrial data. Unlike architecture-led task listings and qualitative frameworks, the approach yields scenario-level quantitative KPIs that capture change propagation across physical, functional, and code layers. A network-graph model encapsulates these indices as scenario-specific state graphs, enabling change tracking and visualisation. The quantification method then computes adaptation effort from graph differences. The approach is validated through adaptation estimates for one lab-scale and three industrial automation cells. Results demonstrate the approach effectiveness in evaluating architecture, implementation, and runtime effort, offering a practical tool for strategic adaptation decisions.

Wireless Sensor Networks (WSNs) have become a foundational technology for real-time monitoring and data acquisition in applications such as environmental surveillance, industrial automation, smart agriculture, and healthcare systems. Despite their widespread deployment, WSNs continue to face significant challenges related to limited energy resources, inefficient clustering mechanisms, scalability constraints, and reduced network lifetime. Traditional clustering and routing approaches are predominantly heuristic or probabilistic in nature and lack the adaptability required to operate effectively under dynamic network conditions. To address these limitations, this research paper presents a deep learning-based framework for cluster optimization in Wireless Sensor Networks, developed from an empirical dissertation study. The proposed framework formulates the clustering process as a supervised multi-class classification problem, enabling intelligent and data-driven cluster formation and cluster head selection. Key network parameters, including residual energy, node distance, node density, and traffic load, are utilized as input features for model learning. A systematic methodology encompassing data preprocessing, neural network-based model development, training, validation, and comprehensive performance evaluation is adopted to ensure robustness and reliability. Experimental results demonstrate that the proposed model achieves an overall classification accuracy of 92.37 percent, with balanced precision, recall, and F1-score values across all cluster categories. Confusion matrix analysis reveals strong diagonal dominance, while training and validation learning curves confirm stable convergence and effective generalization without significant overfitting. The findings highlight the effectiveness of deep learning in enabling adaptive, energy-efficient, and scalable cluster optimization for modern Wireless Sensor Networks.

This paper presents the design, hardware implementation, and performance evaluation of a Brushless DC (BLDC) motor speed control system based on a Variable Frequency Drive (VFD) technique. Conventional speed control approaches relying on constant DC-link voltage with high-frequency PWM suffer from limited dynamic response and poor efficiency under varying load conditions. The proposed system employs an Arduino UNO (ATMega328P) microcontroller to generate variablefrequency six-step commutation signals, which drive a three-phase MOSFET inverter bridge operating in 120° conduction mode. A potentiometer-based reference interface allows smooth, real-time speed adjustment, while an infrared optical speed sensor provides closed-loop feedback. The switching frequency and corresponding motor RPM are displayed on a 16×2 LCD in real time. Hardware test results demonstrate stable closed-loop speed regulation with a measured output frequency of 50.51 Hz and a motor speed of 77 RPM using a 24 V, 250 W, 500 RPM BLDC motor prototype. The system achieves fine-tuned speed regulation, dynamic response, and load-disturbance rejection, making it suitable for electric vehicles, agricultural machinery, and industrial automation

With the continuous deepening of digital transformation, manufacturing enterprises have put forward higher requirements for production collaboration, real-time response and fine control. Focusing on the construction and implementation path of the digital factory under the framework of high efficiency industrial automation system integration, this paper constructs an overall method covering target decomposition, system architecture design, multi-level data integration, collaborative control of industrial Internet and edge computing, digital twin and intelligent analysis and optimization, and phased promotion. In the verification of a discrete manufacturing workshop, 68 key devices and 214 sensing acquisition points are connected to the system. The success rate of device access is 98.6%, the average response time delay is 0.75 s, the success rate of scheduling is 96.0%, the first pass rate of products is increased to 97.1%, and the energy consumption per unit output value is reduced by 18.3%. The research shows that the path can enhance the ability of factory data penetration, business interaction and continuous optimization, and has practical reference significance for the digital upgrading of manufacturing enterprises.

Machine vision and artificial intelligence have reshaped quality inspection across modern manufacturing. In this project, we designed a non-contact metrology system that uses machine vision to inspect mechanical parts like bolts, nuts, and pinions. The setup relies on a high-resolution camera along with a controlled backlight to grab crisp, clear images of components. We run these images through image processing tools and machine learning algorithms to pull out accurate measurements and spot any defects. This system not only boosts accuracy and consistency, but it also runs in real time and slashes the need for manual labor. Our tests show it’s faster and more precise than older, manual methods. It's the kind of upgrade fit for Industry 4.0 and advanced industrial automation.

Wireless sensor networks (WSNs) are the backbone of IoT-enabled smart manufacturing, environmental monitoring, and industrial automation. However, their broadcast nature makes communication links vulnerable to eavesdropping, routing manipulation, and denial-of-service attacks. Strategically placing monitor nodes to check each link is an effective approach to protect against attacks, but energy, connectivity, and capacity constraints should be considered while picking monitor nodes. In this paper, we tackle the Minimum-Weighted Connected Capacitated Vertex Cover (MWCCVC) problem, which minimizes monitoring costs, ensures backbone connectivity, and adheres to per-node capacity constraints. Unlike prior works that consider weighted vertex cover, connectivity constraints, or capacitated variants separately, the proposed MWCCVC model jointly integrates all three dimensions within a single vertex cover-based monitoring framework. We first provide a Branch-and-Bound (B&B) solver with linear programming relaxation bounds and constraint-based pruning strategies that produces optimum solutions. Three constructive greedy heuristics (GD, GR, GW) and two hybrid genetic algorithms (HGA, HGA-v2) that combine parameterized greedy decoders with evolutionary search are proposed; all methods guarantee full edge coverage, induced-subgraph connectivity, and max-flow-validated capacity feasibility. Tests on 130 small, 160 medium, and 19 large benchmark instances show that HGA matches B&B optima on every small instance, beats the time-limited B&B by 6.6% on medium instances, where the percentage is computed based on the relative difference in average total weight with respect to B&B, and stays the best on large graphs with up to 1000 nodes. The HGA-v2 tries to balance the quality and speed, with only a 3.1% difference at faster execution.

ABSTRACT Wireless sensor networks (WSNs) are commonly utilized in various application areas, ranging from environmental monitoring to industrial automation. In WSNs, energy efficiency (EE) is a crucial feature due to sensor nodes' (SNs) limited power resources. Unequal clustering (UC) and routing techniques have emerged as effective approaches to balance energy consumption (ECON) and prolong network lifetime (NLT). However, finding optimal solutions for UC and routing optimization problems in large‐scale WSNs remains challenging. Recently, metaheuristic algorithms have been widely employed to determine the appropriate cluster sizes, cluster heads (CHs), and routing paths that minimize ECON imbalances while maximizing NLT or throughput using various parameters and objectives. Therefore, this study proposes a new bonobo optimization algorithm based on UC with a pelican optimization‐based routing (BOAUC‐POR) technique for energy‐efficient WSNs. The projected BOAUC‐POR technique involves two significant phases: UC construction and routing. In the preliminary phase, the BOAUC‐POR technique uses the BOAUC technique to explore the search space effectively for optimal CH selection and cluster size, considering factors like node density, residual energy (RE), and distance. Besides, the BOAUC‐POR technique follows the POR model for the optimum selection of routes to BS using RE, node degree (ND), and distance. The proposed BOAUC‐POR technique effectively maximizes the EE and NLT in WSN. The simulation values stated that the BOAUC‐POR technique accomplishes superior performance in terms of EE and NLT. The BOAUC‐POR technique exhibits promising results for enhancing the EE and prolonging the lifetime of WSNs, thus enabling sustainable and reliable operation in resource‐constrained environments. The comparison study of the BOAUC‐POR technique portrayed superior values of 510 for HND in the 500‐node scenario and 124 s for CT at 700 rounds.