Journal of Techniques

This study presents an intelligent adaptive PID controller based on Radial Basis Function neural network (PID–RBF) for motion control of a ship subjected to surge, heave, and pitch dynamics. The proposed controller adjusts PID parameters based on tracking errors to perfectly follow the desired trajectory and compensate for nonlinearity and uncertainty. The numerical simulations show that the proposed adaptive PID–RBF controller outperforms the traditional adaptive PID control in terms of accuracy and dynamic performance. The control efforts required by the proposed method have been considerably reduced. In the surge channel, the control energy is reduced from the value 9.58×105to 2.61×104, while there is a reduction in control energy from 3.95×106to 1.22×106 in the sense of pitch channel motion. In addition, the control signal from the proposed PID-RBF controller is smoother than that from the conventional control approach. This indicates that the proposed controller could reduce stress on the actuator and thereby keep it in a safe operating range by lowering the maximum control signal input. In addition, the results show that the proposed controller has better rejection capability and robustness than the conventional controller. The proposed PID–RBF controller demonstrates promising performance in the motion control of a ship across three motion channels (surge, heave, and pitch).

One of the most crucial parts of the oil refining process is the Fluid Catalytic Cracking Unit (FCCU). This system is essential for transforming heavy hydrocarbon fractions into valuable lighter products such as propylene, diesel, and gasoline. This method helps to remove impurities, enhance the quality of finished goods, and increase the yield of desirable products. The unit encounters uncertainties in its model dynamics, cross-coupling between its axes and sub-units, and extremely nonlinear effects. This study discusses the design of four hybrid controllers comprising two PD-Like Interval Type-2 (PDLIT2) fuzzy logic controllers and two Nonlinear Proportional-Integral-Derivative (NPID) controllers for controlling the reactor and regenerator temperatures in FCCU processes. The parameters of the NPID controller and each Fuzzy controller are tuned using the Firefly Optimization algorithm. These controllers make up for the model's nonlinearities, uncertainty, chattering, and cross-coupling. The overshoot of the reactor temperature response is enhanced by up to 85%, while that of the generator is enhanced by up to 60%, compared with previous published work.

In this work, a novel voltage control method for DC–DC buck converters using an artificial neural network (ANN) is proposed under nonlinear, time-varying load conditions. Traditional PID controllers typically exhibit poor performance in response to abrupt load variations and parameter uncertainties. To overcome these limitations, a lightweight feedforward ANN controller is formulated and trained with data from the MATLAB/Simulink buck-converter model under diverse operating scenarios. For the model itself, the proposed controller provides direct estimation of the optimal pulse-width modulation (PWM) duty cycle, enabling accurate voltage regulation and rapid transient response without an explicit mathematical model. The simulation results show that the ANN-based controller exhibits lower overshoot, shorter settling time, and better load and line regulation than a conventional PID controller. Also, the controller is very resilient to component variations or external disturbances. In contrast to ANN–model predictive control (ANN–MPC) hybrid architectures described in the literature, the current proposed approach does not require predictive optimization or expert-based training. It is thus less computationally demanding and more suitable for real-time embedded and FPGA-based applications. The proposed method yields rapid, stable voltage regulation during transient simulations under both start-up and load disturbance conditions

In this study, three lightweight classifiers for a four-class epilepsy seizure type described by 16-channel EEG features have been benchmarked. Models have been evaluated on identical splits using standard accuracy, error rate, training time, prediction throughput and memory footprint, and then, further inspected through scatter-plot diagnostics, confusion matrices, ROC curves, and post-hoc explainability (SHAP and LIME). All algorithms achieve high validation accuracy (≥ 95 %), yet exhibit distinct trade-offs. The ensemble KNN delivers the best overall fidelity (98.8 % validation, 99.3 % test; macro-AUC ≈ 0.9996) but requires the largest model (25 MB) and provides only mid-range inference speed (3.8 k obs/s). The SVM matches the KNN's test accuracy and is the most balanced option, training in the quickest time (14s) and occupying < 2 MB. The ANN trails slightly in accuracy (97.1% test) but is two orders of magnitude smaller (24 kB) and fifty-fold faster at prediction (181 k obs/s), thus it is better suited for edge or real-time deployments. SHAP and LIME analyses converge on a common set of influential variables, primarily X15 and X16 with contributions from X8 or X11/12, strengthening confidence in the feature representation and suggesting a reduced sensor subset for future work.

Blockchain applications have adopted various techniques over the past decade, and one revolutionary approach has been the use of Artificial Intelligence (AI). They had crucial roles in the transparency and security of e-Governance applications. In this paper, the interrelated impacts of AI techniques in enhancing Blockchain applications are illustrated to ensure transparent and secure e-Governance services against forgery and intrusion. By reviewing available articles, this paper discusses the role and capabilities of AI in influencing and manipulating large volumes of data related to Blockchains, which build stubborn, robust, and stable e-governance. The conclusions of this paper reveal that adopting AI techniques in Blockchain applications significantly enhances security and transparency by enabling simultaneous detection of system abnormalities. Concluded outcomes highlight major expectations for Blockchain AI Combination (BAIC) techniques to enhance e-government frameworks and financial transactions, emphasizing their potential to support global achievements by promoting greater transparency and security in economic environments. In this review, deeper and broader research into the scalability of Blockchain combined with AI techniques is proposed to address various aspects of e-governance.

Natural language generation systems sometimes struggle to model long-range semantic connections and maintain contextual consistency, especially when applied to linguistically sophisticated literary corpora. Traditional recurrent neural architectures are good at modeling sequential patterns but typically fail to preserve higher-level thematic and stylistic information in text production. The current paper proposes a semantic-aware hybrid framework based on Word2Vec embedding representations, ++K-Means semantic clustering, and Bidirectional Long Short-Term Memory (Bi-LSTM) sequence learning to improve contextual coherence and next-word prediction performance. In the proposed architecture, the model learns to obtain semantic context vectors from clustered embedding spaces and to fuse them with sequential hidden representations for better language modeling. This study examines the system on three benchmark datasets from English corpora of both literary and general domains: the Nietzsche corpus, Shakespeare plays and WikiText-2. Experimental results show that the proposed semantic-aware recurrent architecture consistently outperforms the standard statistical and neural baseline models. The model achieves prediction accuracies of 67.4%, 61.3%, and 63.1% on the Nietzsche, Shakespeare, and WikiText-2 datasets, respectively, while reducing perplexity values and enhancing linguistic coherence. A more detailed analysis of the robustness test, semantic error evaluation, and ablation experiments confirm that semantic clustering effectively can improve contextual consistency, stylistic preservation, and semantic continuity. The results demonstrate that combining clustering-based semantic abstractions with recurrent sequence modeling is an effective, computationally lightweight approach to context-aware text synthesis for both literary and general-domain applications.

This paper describes the design, evaluation, and investigational validation of a dual-band Substrate Integrated Waveguide (SIW) supported Dielectric Resonator Antenna (DRA) for fifth-generation (5G) applications. The designed antenna is realized on a Rogers RT/duroid 5880 substrate (εr = 2.2, tanδ = 0.003, thickness = 1.2 mm) and excited by a microstrip-fed SIW structure. A cubic Dielectric Resonator (DR) with dimensions of 5.2 × 5.2 × 5.2 mm³ and dielectric constant of 9.9 is positioned over a modified H-shaped slot engraved on the SIW cavity. The proposed modified-H-shaped slot serves as a magnetic-coupling aperture, enabling controlled excitation of multiple resonant modes and facilitating dual-band operation at 28 GHz/38 GHz in the Ka band. The designed antenna achieves a measured impedance bandwidth of 12.12% (26.02–30.26 GHz), a realized gain of 7.1 dBi, and a radiation efficiency of 84.2% at 28 GHz. When operated at 38 GHz, it displays a bandwidth of 8.95% (37.01–40.07 GHz), a maximum gain of 7.28 dBi, and a radiation efficiency of 81.9%. Moreover, the proposed antenna exhibits circular polarization with an axial ratio below 3 dB at both operating frequencies, enabled by the combined contributions of the altered H-slot and the dielectric resonator. The antenna is constructed and optimized using a full-wave electromagnetic simulator (CST Microwave Studio), and its performance is well-matched between simulation and measurement results. The new SIW-DRA structure is seen to improve several parameters as compared to the SIW slot antennas.

The deep-drawing process for square cups is primarily used in sheet metal production, but it is associated with several defects, such as tearing, earing, and wrinkling. This research intends to optimize the square deep drawing process to minimize or eliminate earing defects using an inverted deep drawing die design and to produce an optimal cup (without defects). Thus, the design and manufacture of the inverted deep-drawing die for the production of square cups are carried out experimentally and by the finite element method using the ANSYS program. The investigation is carried out by conducting several experiments using three metals (low carbon steel st1008, brass CuZn30, and aluminum A1100). The metal thickness is 0.7mm, and the materials are characterized by tensile testing and mechanical properties, including Young's modulus and yield stress. Experimental tests have been carried out under various load conditions for the empty holder. A three-dimensional square-cup model (43mm by 43mm) with a (0.7mm) thickness is built. Given the sample's symmetrical geometry, only one-fourth of the model needed to be analyzed. The minimum Blank Holding Force (BHF) for steel and brass sheets is higher than that required for aluminum sheets. The percentage difference between the experimental punch force and ANSYS test results is 7.4% for steel, 9% for brass, and 6.5% for aluminum. An important thing to note during the formation of square cups is that the deformation conditions in the die's hollow change. In fact, the corners of the cup are subject to greater deformation than the sides, because during distortion the metal on the sides of the cup moves more uniformly and more easily than in the corners.

The recent advancements in laser welding (LW) technology through the incorporation of Digital Twin (DT) technologies, due to the growing importance of Industry 4.0 and smart manufacturing in the context of automation, precision, and digital fabrication, have made LW one of the world's leading advanced production processes. The DT technologies are expected to provide the necessary integration to facilitate effective real-time monitoring, predictive modeling of process behavior, optimal process control, and adaptive process control capability for welded parts. However, an overall framework for effectively integrating LW technology into the manufacturing environment has yet to be developed. This paper critically reviews recent developments in DT-enabled LW systems and proposes a hybrid framework for successful industrial deployment. For data collection, the authors used a systematic review approach to identify studies published after 2018 that covered various technologies related to DTs applied to welding. The review included consideration of real-time sensing technologies; data fusion using multi-sensor technologies; physically based modeling technologies; artificial intelligence technologies; EDGE-Cloud Computing architectures; and adaptive process control strategies. The findings indicate that developing a physics-informed simulation of the welding process and integrating it with an AI-based prediction model will significantly improve accuracy in assessing weld quality, predicting defect occurrence, and optimizing the process. The integration of EDGE Computing and Cloud Computing with Office & Manufacturing processes supports real-time responses and scalability within the industry. Some of the major challenges identified in the research included data synchronization, interoperability, cybersecurity, computational latency, and explainable artificial intelligence. A hybrid DT framework is proposed to provide an intelligent, autonomous platform for future LW applications in Smart Manufacturing.

This paper aims to evaluate how Fe-doped ZnO nanorods with controlled oxygen-vacancy levels can improve random lasing performance for photonic applications. An investigation of the electronic structure, defect dynamics, and optical properties of these nanostructures is accomplished through density functional theory (DFT+U) calculations. Specifically, it has been reported that the incorporation of about 3 at.% Fe can introduce intermediate 3d energy levels in the ZnO bandgap, suggesting the possibility of carrier population inversion and an estimated reduction of the lasing threshold by approximately 30–40% for Fe-doped ZnO nanorods compared to undoped nanorods, under idealized theoretical assumptions. It is also determined that oxygen vacancies act as scattering centers, altering the local electromagnetic environment around the ZnO nanorods and thereby optimizing the gain. Additionally, quantum confinement is predicted to dominate the emission from ZnO nanorods with diameters below 4 nm, thereby shifting the emission spectrum to higher energy (i.e., a blue shift) and increasing the nanorods' oscillator strength. Therefore, it is fair to anticipate that optimum random lasing would occur in ZnO nanorods having diameters of 3-4 nm, doped at 2-3 at.% Fe, and synthesized in oxygen-deficient environments. Although all computations in this study have made assumptions about the configuration of defect sites and the degree of oxidation of ZnO nanorod surfaces, these theoretical results may help to develop the next-generation materials for a range of applications, including biosensing, imaging, and photonics.