The automation of material receiving processes plays a central role in ensuring traceability, process control, and data consistency in warehouse and logistics environments. Inbound material registration is a critical stage, as inaccuracies at this point propagate throughout subsequent operational processes. This paper presents an integrated automation framework focused on the monitoring and control of material entry processes, emphasizing structured data acquisition and system-level control. The proposed framework is composed of a mobile application designed for material intake registration and a centralized web-based platform responsible for data storage, validation, and monitoring. The inbound application supports the capture of material information through document scanning, image recording, and structured input fields, while the web platform consolidates records, enables historical tracking, and manages access and device control. From a methodological perspective, the system architecture was developed based on automation and control principles, with emphasis on modularity, process standardization, and consistency of information flow. The framework enables real-time registration of inbound events and continuous monitoring of material entry status through a centralized interface. Experimental deployment in a controlled warehouse scenario demonstrated the system’s ability to ensure coherent registration of inbound materials, reduce information gaps during receiving operations, and support traceability through structured data records. The results indicate that the proposed framework provides a consistent and controllable approach to inbound material monitoring. From a theoretical standpoint, the study contributes to the understanding of automation-oriented system architectures for material entry processes, offering a reference model for future developments in warehouse automation and control systems.

The integration of legacy industrial assets into IIoT frameworks requires standardized protocols compatible with resource-constrained hardware. This paper evaluates an implementation of the OPC UA PubSub protocol (IEC 62541-14) developed in MicroPython for an ESP32 microcontroller, addressing the challenge of deploying industrial communication standards on low-cost embedded platforms. This research benchmarks the protocol stack's performance through systematic 50-sample statistical analysis, comparing local and cloud MQTT broker deployments. The methodology isolates protocol processing overhead from network infrastructure effects, enabling the precise characterization of implementation efficiency. The results demonstrate a deterministic internal serialization time of 8.46 ms (± 1.26 ms), which remains invariant regardless of network infrastructure, validating the protocol's computational stability. The library maintains a minimal memory footprint, with a peak consumption of only 4.86 KB per message cycle. Network performance analysis revealed significant differences between deployment scenarios. Local broker deployment achieved an average RTT of 119.71 ms with a jitter of 111.37 ms, while cloud broker deployment exhibited an RTT of 985.36 ms with a jitter of 206.49 ms. Despite the 8.2× latency increase in cloud scenarios, the system maintained a 100% message delivery success rate without packet loss across all test conditions. This study concludes that this implementation provides a robust foundation for industrial telemetry and monitoring applications in brownfield scenarios, demonstrating resilience under varying network conditions while maintaining strict protocol compliance with IEC 62541-14.

This work proposes an embedded data infrastructure based on ESP32 microcontrollers integrated with ESP32-CAM modules, focusing on the construction, management, and utilization of a centralized facial image database for emotion-based student readiness assessment. In the proposed architecture, multiple ESP32-CAM devices operate as distributed embedded sensing nodes, capturing facial images of students during academic assessment activities. Each embedded node performs lightweight preprocessing tasks, including image resizing and noise reduction, to optimize transmission efficiency and reduce network load. The preprocessed images are transmitted via Wi-Fi to a centralized database system, where they are securely stored and indexed for subsequent analysis. The database serves as a core component for large-scale data organization, enabling automated processing, historical tracking, and statistical aggregation of emotional data. A convolutional neural network (CNN), trained on the FER2013 dataset, analyzes the stored images to infer facial emotion categories, which are subsequently mapped to quantitative indicators of concentration and nervousness. This separation of embedded data acquisition from centralized analysis allows improved scalability, efficient resource utilization, and flexibility for future model updates without modifications to the embedded hardware. Experimental results demonstrate that the ESP32-CAM platform provides reliable long-term operation and consistent image quality in classroom-like environments. The proposed architecture highlights the role of embedded systems not only as data acquisition devices but as fundamental components of data-centric, emotion-aware educational platforms.

Embedded automation and control systems increasingly depend on continuous monitoring of environmental variables, particularly temperature and relative humidity, under strict energy and computational constraints. Recent advances in Tiny Machine Learning (TinyML) enable predictive models to be executed directly on microcontrollers, requiring explicit trade-offs between predictive accuracy, memory footprint, execution latency, energy consumption, and operational robustness. This work presents a comparative evaluation of three lightweight neural network architectures—a multilayer perceptron (MLP), a one-dimensional convolutional neural network (Conv1D Tiny), and a long short-term memory network (LSTM)—implemented on an ESP32 microcontroller for temperature and humidity time-series modeling. Two execution scenarios are investigated, in which both replay and field modes employ the same on-device rolling window composed of 24 valid samples. In replay mode, deterministic input data are used as a deterministic test bench for controlled validation. In field mode, the rolling window advances as new sensor samples are acquired during real operation. Experiments were conducted using an offline evaluation workflow, referred to as LiteML-Edge, employed as an experimental tool for model training, testing, and consistency checks between offline evaluation and on-device execution. Model performance is assessed using energy-aware deployment-orientated criteria central to control systems, including inference latency, flash and RAM utilization, and energy-related measurements, together with standard regression metrics such as mean absolute error (MAE), root mean squared error (RMSE), and coefficient of determination (R²). Results indicate that the LSTM achieves higher predictive accuracy under controlled replay conditions, while the MLP demonstrates higher robustness and lower computational overhead during field operation. The Conv1D Tiny model exhibits intermediate behavior, balancing limited temporal modeling capability with moderate memory usage and energy efficiency. These results confirm that no single architecture is universally optimal and that model selection should be guided by execution context and control constraints.

Hybrid manufacturing has emerged as an advanced manufacturing technique to fabricate complex, critical components with precision and ensure functional performance. Such technologies face some obstacles, such as process variability, defect formation and insufficient real-time adaptiveness. This study proposes an intelligent framework to mitigate such limitations by incorporating an ML-driven real-time monitoring tool for process control, including defect identification and quality optimisation. This approach integrates multi-sensor in situ monitoring tools during the fabrication process, such as temperature sensors, visual sensor data and process-related vibration data. A convolutional neural network (CNN) is utilised to identify spatial attributes associated with surface characteristics and defect patterns during the process. A long short-term memory (LSTM) network is employed to capture time-dependent relationships within process signals that are combined with a CNN to ensure in situ defect identification and predict process quality status, including surface integrity and trends in mechanical properties. Handcrafted statistical feature extraction and smart anomaly-driven image features inspired by activity recognition are utilised in the LSTM network to identify predefined types of anomalies. This ML-driven framework develops an adaptive control strategy to execute the real-time opitimisation of critical parameters, such as energy consumption, feed rate, heat input and tool path. Therefore, such an intelligent data-driven approach ensures defect mitigation and stabilises the process by facilitating the advanced closed-loop decision-making in hybrid manufacturing environments. Higher accuracy ensures the model's capability for in situ process monitoring and defect identification. Moreover, the proposed framework also achieves superior performance compared to conventional ML approaches. Extensive robustness checks of this proposed CNN–LSTM framework are required to adopt and implement it for large-scale industrial applications.

Adhesive wear is a dominant problem for copper in electrical sliding contacts due to the low relative hardness difference between the contacting bodies materials and their high mutual solubility. These adhesive islands are responsible for intermittent electrical contact during sliding, eventually leading to a loss of reliability and performance in electromechanical systems such as electric drives, slip-ring assemblies and current collectors. To mitigate this issue, a solid lubricant reinforcement approach and powder metallurgy route were adopted. Layered materials were reinforced into the copper matrix and processed using the spark plasma sintering technique. The processed composites were tested under electrical sliding conditions. The obtained results were statistically analysed using response surface methodology to identify the most influential parameters affecting tribological performance. This was followed by worn surface analysis, which revealed reinforcement-specific wear signatures. The influence of electric current was evident by the dominance of electrical arc erosion mechanisms. This research addresses the combined effects of electric current and mechanical load on the coefficient of friction, electrical contact resistance, and wear as key performance indicators for different composites. The stability of electrical contact resistance was found to be strongly influenced by the size of wear debris. SEM-based debris analysis revealed molten and sintered debris on the wear tracks, which were subsequently oxidised under ambient air conditions.

Introduction: The most common processes for the production of multifunctional composite sandwich panels include vacuum-assisted hand lay-up, vacuum infusion, among others. However, these processes are very human-dependent, especially vacuum-assisted hand lay-up. This work aims to develop, optimize, and validate a controlled vacuum infusion manufacturing process for multifunctional composite sandwich panels, enabling scalable, efficient, and high-quality production while improving mechanical performance and process consistency.

Methods: The developed equipment incorporates dual heated plates operating at temperatures up to 200 °C, segmented thermal zones, and a vertical actuation system to ensure uniform pressure during processing. It enables the manufacture of panels measuring up to 1.0 m × 2.0 m × 0.2 m and can be reconfigured into a 2.0 m × 2.0 m working surface through a rotatable upper plate, enhancing operational flexibility. This configuration allows for the single-step production of sandwich panels via vacuum infusion, significantly reducing production time. Curing cycles were optimized using Differential Scanning Calorimetry (DSC) in accordance with ASTM D3418, considering the limitations of the raw materials. Sandwich panels incorporating glass or basalt fibres, an epoxy matrix, and an extruded polystyrene (XPS) core were produced and evaluated. A comprehensive experimental campaign, including DSC, Barcol hardness, density, fibre and void content, flexural performance, and dimensional stability tests, was conducted following relevant international standards.

Results: The results indicate that sandwich panels manufactured with basalt fibres using the vacuum infusion process exhibit superior overall performance. Additionally, vacuum infusion enables faster and more consistent production compared to vacuum-assisted hand lay-up.

Conclusions: The validated infusion strategy and control system demonstrate strong potential for scalable manufacturing, with further work planned to complete performance validation and refine process optimization.

In the automotive industry, ensuring continuous production and maintaining equipment reliability are vital for meeting both quality expectations and delivery schedules. Radiator manufacturing relies heavily on core builder machines, which are responsible for assembling radiator cores with precision and consistency. Any unplanned downtime in these machines can significantly disrupt the manufacturing process, leading to production delays, increased operational costs, and potential risks to product quality.

A common challenge encountered with core builder machines is the frequent failure of their motion systems, particularly those exposed to high cycle rates and harsh working environments. For example, in one case, a belt-driven servo piston system used in a core builder machine was found to fail approximately every two months. These failures resulted in notable unplanned downtime, impacting overall production efficiency and placing additional strain on maintenance teams.

To address this issue, a practical maintenance improvement was implemented by replacing the original belt-driven servo piston system with a pneumatic piston solution. This modification aimed to enhance the reliability of the machine, reduce the frequency of unexpected breakdowns, and improve maintenance efficiency. After the change, operational data and maintenance records were reviewed to evaluate the impact on machine performance. The observations revealed that the pneumatic piston system performed more consistently under demanding conditions, leading to fewer failures, reduced downtime, and lower maintenance costs.

Overall, this study demonstrates that targeted maintenance improvements can have a positive impact on equipment reliability and production continuity in radiator manufacturing.

Flux-modulating machines are emerging as noteworthy machines, particularly in wind turbines. They typically have two windings on their stators, which are not directly coupled, but use a flux-modulating rotor to cross-couple the two stator windings. These brushless machines are typically medium-speed machines, and they boast reliability suitable for remote areas with low accessibility.

Two popular machines in this category are the brushless doubly fed machines (BDFMs) and wound field flux switching machines (WFFSMs). Although these machines work according to similar operating principles, they have almost contrasting descriptions. BDFMs, viewed as alternatives to doubly fed induction generators (DFIGs), are noted for their comparably lower power density and efficiency. Conversely, WFFSMs are typically touted for their high power densities, hence the suggestions of replacing PM machines.

In this paper, the performances of BDFMs and WFFSMs are compared. Select topologies that operate at the same speed are evaluated: the 4/6 BDFM and the 24/10 WFFSM. Instead of applying a direct comparison between these two topologies, optimized 250 kW designs are compared to their conventional parallel topology. BDFMs are compared with DFIGs, while WFFSMs are compared with wound field synchronous machines. The optimization processes are conducted using the non-dominated sorting genetic algorithm coupled with response surface approximations from FEA evaluated designs.

It is shown that a 4/6, 250 kW BDFM typically has about 5 % less efficiency compared to a 20-pole DFIG, while being almost two times the volume. The efficiency in WFFSMs is better due to less copper loss; however, their power factors are low. This is due to the cross-coupling nature of operations in BDFMs and WFFSMs, which leads to significantly lower flux utilization compared to their directly coupled counterparts. However, their reliability advantages cannot be dismissed.

This paper proposes an online maximum torque per ampere (MTPA)-based current control and a modified speed estimation algorithm using a full-order speed observer for field-oriented control (FOC) of a synchronous reluctance motor (SynRM) traction drive for EVs. Unlike conventional MTPA methods based on offline look-up tables, the proposed approach computes the d-axis current reference online using a current-optimizing factor. The key concept is that the current-optimizing factor adapts automatically with operating conditions to keep the excitation level optimal: when load torque increases, the flux-producing component is increased to support torque production; however, during transient events, the current-optimizing factor decreases automatically, which limits the rise in flux and current and prevents over-excitation. At low speed or light load, the current-optimizing factor maintains an optimal stator current, reducing losses and improving overall efficiency. Quantitatively, although the conventional method also maintains the operating point, the proposed optimization reduces the flux from about 1.00 pu to about 0.85 pu at nearly nominal load (0.9 pu), while improving the power factor from about 0.45 to 0.65 at low load (0.2 pu) and from about 0.75 to 0.87 at high load (0.9 pu) in both motoring and regenerative modes. Additionally, to improve the robustness of speed estimation, the full-order observer is augmented with an explicit stabilization function. The stabilization function is obtained from the dot product of stator current and flux and is used to minimize oscillations and instability during low-speed motoring and regenerative modes, including under parameter variations and transients. The proposed speed estimator is evaluated by comparing the estimated speed with the motor speed across several EV driving profiles, including motoring, constant-speed, low- and zero-speed, and regenerative modes. The results show stable operation under load–torque disturbances and accurate speed tracking across the tested profiles, demonstrating that the proposed online current-optimizing MTPA strategy and stabilization-function-enhanced full-order observer are effective for sensorless SynRM traction control.