Phthalates (PAEs) are a group of chemicals widely used as plasticizers in various industrial and consumer products. Due to their extensive use and potentially harmful effects on human health, particularly as endocrine-disrupting compounds, the rapid and efficient detection of PAEs is crucial. Recent developments in sensor technology have significantly enhanced the monitoring of environmental and food contaminants. This study aimed to develop a surface plasmon resonance (SPR) sensor utilizing diethyl phthalate-imprinted nanoparticles (DEP-MINPs) for the sensitive and selective detection of DEP. DEP-MINPs were synthesized via microemulsion polymerization and immobilized onto a bare gold SPR chip as a recognition layer. The DEP-MINPs were characterized using FTIR, TEM, and zeta potential analyses. The kinetic performance was investigated within a concentration range of 1.0–135.0 μM, observing a strong linear correlation between DEP concentration and SPR sensor response. The calculated limit of detection was 0.30 μM. The Langmuir model best described the interaction, suggesting monolayer adsorption on a homogeneous surface. Selectivity studies using structurally similar compounds, including dimethyl phthalate, styrene, and vanillic acid, revealed that the sensor was 2.64, 5.00, and 4.83 times more selective for DEP, confirming the success of the molecular imprinting process. The developed SPR sensor was successfully used for DEP analysis in bottled water and sanitary pad samples. The DEP-MINPs SPR sensor demonstrated excellent reproducibility and stability over multiple measurements and days, supporting its potential for practical applications in environmental monitoring.
The inherent stochasticity of wind resources poses fundamental challenges to the reliable integration of wind power plants (WPPs) into modern grid infrastructures. This study introduces a novel, integrated framework that synthesizes high-fidelity, empirically calibrated digital twin modeling—built upon the National Renewable Energy Laboratory’s System Advisor Model (NREL SAM) using validated operational data from existing WPPs—with statistical process control (SPC) methodology to establish a robust, multi-scale early warning system for wind energy systems. A comprehensive, empirically calibrated digital twin model of a 48 MW WPP, comprising 32 GE 1.5 SLE turbines situated in Colorado, USA, was constructed using the NREL SAM platform, which is extensively validated against real-world operational data from existing wind power plants. The model incorporated site-specific meteorological parameters, detailed turbine specifications, rectangular farm configuration with eight rotor diameter spacing, and an extensive loss taxonomy yielding an 11.02
To characterize the deformation irreversibility and loading–unloading path dependence in aeolian sand–loess mixtures (ASLMs), one-dimensional confined loading–unloading tests were conducted on specimens with sand mixing ratios of 0
Vibration-based structural health monitoring (SHM) has become an important tool for damage detection, condition assessment, and integrity evaluation of engineering structures. Consequently, the variational iteration method (VIM) has attracted significant attention due to its accuracy and efficiency in the free vibration analysis of beam structures. However, its application to complex configurations, such as stepped, multi-segment, or multi-cracked beams, has remained limited, primarily because of the high computational cost and memory requirements associated with extensive symbolic integrations. To overcome these limitations, the present study employs an enhanced numeric–symbolic hybrid variational iteration method (VIM), originally developed in earlier foundational studies and recently extended by the authors. This hybrid formulation significantly reduces computational time and memory requirements when applied to complex vibration-based damage assessment problems. In this work, the proposed methodology is further extended to analyze the dynamic behavior of generalized multi-span cracked Timoshenko beam systems, including beams with cracks repaired using bonding (filling) materials. To the authors’ knowledge, this represents the first application of the hybrid VIM framework to the vibration analysis of beams containing filled cracks. The proposed model is validated by comparison with existing analytical solutions for uncracked and unfilled cracked beams. To capture the vibration characteristics of the filled crack, an independent finite element model is developed in ANSYS. The results demonstrate an excellent agreement between the present formulation and the reference solutions, confirming the accuracy and robustness of the proposed approach. A parametric study is subsequently conducted to examine the crack-healing phenomenon via crack filling, with particular emphasis on the effect of the filler material’s shear modulus relative to that of the host beam. The results reveal that, under the assumption of complete filling and perfect bonding between the filler and host materials, even a filler material with a relatively low elastic modulus can effectively restore the stiffness and vibration characteristics of the damaged region, closely approximating the dynamic response of a structurally healed beam.
This paper presents a fracture-mechanics-based analytical framework for evaluating the virtual crack closure integral (VCCI) of structures containing filled cracks. Building on the formulation of Fowkes et al., closed-form expressions for the stress and displacement fields in the vicinity of a filled crack are derived by explicitly accounting for the presence of a wedge-shaped filling material. Using these fields, the virtual crack closure integral is analytically evaluated for the filled-crack problem. Finite element simulations are also conducted using ANSYS to evaluate the J-integral at the crack tip, which is equivalent to the VCCI. Comparisons between the analytical and numerical results for different filling material properties and crack depths show excellent agreement, validating the proposed formulation. The influence of the wedge angle on VCCI is examined. The applicability of the stress intensity factor to filled cracks is also discussed, and its limitations and relationship to energy-based fracture parameters are clarified. In addition, the effects of Poisson’s ratio of both the filling material and the surrounding matrix are investigated. The proposed model provides an efficient tool for evaluating fracture parameters in filled-crack systems and can be readily incorporated into simplified static and dynamic structural analyses, thereby supporting the development of more reliable and resilient structures and contributing to infrastructure sustainability.
Solar-driven photothermal interface evaporation technology offers an attractive route to mitigate water scarcity with low environmental burden and favorable economics, yet it remains nontrivial to engineer evaporators that simultaneously deliver high flux, operational stability and apply across different real-world conditions. Herein, a three-dimensional (3D) hierarchical porous cement-based architecture was constructed, with carbon dots incorporated into interfacial regions to form a carbon-dot-enhanced cement-based evaporator (CDs-CBE). The uniformly distributed carbon dots serve as reinforcing constituents, significantly enhancing broadband light absorption and photothermal conversion performance. The optimized CDs-CBE−2 achieved excellent evaporation rate of 2.16 kg m−2 h−1 and evaporation efficiency of 91.6
Metal additive manufacturing (MAM) has transformed the creation of complex geometries for the aerospace, energy, and biomedical industries, but widespread industrial use has been limited by inherent process limitations such as poor surface finish, dimensional error, and critical defects such as porosity and residual stresses. This review thus presents a thorough analysis of in situ hybrid additive manufacturing (HAM) systems as a practical pathway to address these problems. This review first categorizes and analyzes the main types of defects associated with fusion-based processes, i.e., powder bed fusion (PBF) and directed energy deposition (DED), such as lack of fusion, keyholing, and solidification cracking. Building on this basis, the paper critically explores emerging in situ hybrid architectures combining additive deposition with secondary processes such as subtractive machining, plastic deformation (e.g., interlayer rolling, peening), and laser-based finishing within a single-chucking environment. The review focuses on the active mitigation of defects, the microstructure refinement, and surface integrity enhancement in real time through synergistic combinations, as opposed to sequential post-processing approaches. Furthermore, the critical role of digital integration is discussed and includes the development of advances in hybrid process planning and multi-modal in situ monitoring and the implementation of artificial intelligence (AI) and machine learning (ML) for closed-loop control. Finally, the paper evaluates the techno-economic feasibility of hybrid routes and finds key challenges that still remain in qualification, standardization, and data infrastructure needed for full industrial deployment.
Experimental investigations were carried out to characterize the flame stability and combustion dynamics of non-premixed LPG–air flames in a swirl-stabilized combustor addressing a notable gap in the literature. Experimental data for non-premixed LPG–air flames, across gas turbine-relevant firing rates and varying burner geometries remain limited. The experiments were conducted over a range of equivalence ratios, from the lean blowout (LBO) up to stoichiometric conditions (Φ ≤ 1), and combustor firing rates between 4 and 10 MW/m3-bar. Two annular air ducts with diameters of 12 mm and 16 mm were employed to vary the oxidizer Reynolds number. The novel contributions of this work are: (1) quantification of LBO equivalence ratios for non-premixed LPG–air flames over firing rates representative of micro-gas turbine operation; (2) demonstration that increasing the annular air duct diameter from 12 to 16 mm extends the lean stability limit by reducing the oxidizer Reynolds number and the associated flame strain rate; and (3) provision of axial temperature profile data suitable for validating numerical models of LPG–air combustion, for which benchmark experimental datasets are scarce. The results indicate that the LBO equivalence ratio is consistently lower for the 16 mm burner compared to the 12 mm configuration across all firing rates investigated. Additionally, the LBO equivalence ratio increases with increasing firing rate, primarily due to elevated turbulent strain rates. Froude number analysis reveals that the flames are buoyancy-dominated at blowout conditions for all firing rates considered. Higher axial temperatures are recorded in the 12 mm burner configuration.
Chemical enhanced oil recovery using nanoparticles and surfactants has attracted the attention of researchers because of its efficacy in improving oil displacement efficiency. This research aims to investigate the impact of Al2O3 nanoparticles on cetyltrimethylammonium bromide (CTAB) cationic surfactant for improving oil recovery in low-permeability carbonate reservoirs. The presence of different mineralogy and functional groups in carbonate rock was analyzed by XRD and FTIR, respectively. A conductivity test and the surface tension method detected the critical micelle concentration (CMC) of CTAB to be around 0.9 mM. The nanofluids (Al2O3-CTAB) were found to be stable for more than a week based on visual observation and the zeta potential values of above + 30 mV. The interfacial tension (IFT) between crude oil and nanofluids was reduced to 5.4–6.7 mN/m. The solid–liquid contact angles were reduced to 28–39°, indicating a favorable water-wet condition. The emulsification of crude oil using nanofluids was observed at 30 °C and at an elevated temperature of 70 °C, highlighting the potential of Al2O3-based nanofluids. Core flooding experiments, representing dynamic behavior, resulted in cumulative oil recoveries of 23.4
Grouting is widely used to mitigate geotechnical hazards in underground engineering. However, conventional cement-based grouts often suffer from shrinkage-induced debonding at the grout–rock interface, thereby compromising reinforcement performance. Expansive grouts can effectively alleviate this problem; however, their expansion effects on grout diffusion and reinforcement have not been adequately investigated. This paper develops a numerical method for simulating expansive grout that incorporates both hydro-mechanical (HM) coupling effects and grout expansion effects. The proposed method is validated against laboratory tests. A three-dimensional numerical model of expansive grouting in a single-fracture rock mass is established to investigate the HM coupling and expansion effects on grout diffusion and rock deformation. The method is further applied to grouting reinforcement for a practical underground petroleum storage cavern project. Results indicate that the effective expansive agent contents above 3
This paper presents a Multi-Objective Differential Evolution framework (MODEA) and an enhanced Multi-Objective Particle Swarm Optimization (MOPSO) technique for power transformer design optimization. The proposed approaches integrate multi-objective optimization mechanisms to improve solution quality, convergence, and diversity of the Pareto-optimal front. A comprehensive comparison is conducted against three widely used Multi-Objective Evolutionary Algorithms (MOEAs), namely the Strength Pareto Evolutionary Algorithm (SPEA), Non-Dominated Sorting Genetic Algorithm II (NSGA-II), and Non-Dominated Sorting Genetic Algorithm III (NSGA-III). The study simultaneously minimizes active-part cost and maximizes transformer efficiency while evaluating convergence, diversity, and Pareto-front quality. Unlike conventional Differential Evolution approaches, MODEA integrates differential mutation with Pareto-based ranking, crowding-distance diversity preservation, and elitist selection mechanisms within a multi-objective optimization framework tailored for constrained transformer design problems. MATLAB-based simulations demonstrate that MODEA consistently produces high-quality and well-distributed Pareto-optimal solutions. Furthermore, the best compromise solution obtained by MODEA achieved an 11.17
Joining dissimilar austenitic stainless steels is difficult because the two alloys respond differently to heat input and plastic deformation, and fusion welding of these grades often produces solidification defects and wide heat-affected regions. The challenge is to select solid-state welding parameters that give a sound joint with balanced strength, ductility, and toughness. This study investigates the effects of rotational speed and feed rate on the thermal response, microstructure, and mechanical performance of rotary friction-welded dissimilar AISI 304/AISI 316 L stainless-steel rods. Joints were produced at rotational speeds of 1000, 1400, and 2000 rpm and feed rates of 40 and 63 mm/min. The welds were characterized by temperature monitoring, optical microscopy, SEM/EDS, microhardness mapping, tensile testing, Charpy impact testing, and fracture surface analysis. Higher rotational speed increased the peak temperature and produced a more uniform weld region with fewer visible interfacial defects. Tensile strength increased with rotational speed for both feed rates, reaching 550 MPa at 2000 rpm and 63 mm/min. Impact toughness also increased with rotational speed, reaching 111 J/cm2 at 2000 rpm and 40 mm/min. Hardness profiles showed clear zone-dependent variations across the joint and slightly higher hardness on the rotating AISI 304 side. Fractography showed a transition from flat, void-containing fracture surfaces at low rotational speed to dimpled, ductile tearing features at high rotational speed. Within the tested parameter window, rotational speed mainly controls joint consolidation, while feed rate influences the balance between strength and toughness, especially at low and intermediate speeds.
This study presents a novel five-level converter (NFLC) architecture for integrating electric vehicle (EV) and photovoltaic (PV)-based micro-grid. The NFLC converter is designed with fewer switches, has reduced converter cost, and can be applied to systems in industries and power plants employing high power and medium-voltage ranges. There are two distinct DC-link connections on the NFLC. A new Bell polynomial-based controller is proposed in this paper for the control of NFLC. The proposed control method regulates almost identical values across the two DC-link voltages of the NFLC. Further, this new converter is connected to a single-phase power distribution system and used as a compensator. The grid, load, solar panel, compensator, and EV are modeled and integrated into the proposed system. The designed system is modeled in MATLAB/SIMULINK. The controller is designed to support the grid, and both charging and discharging the EV battery are feasible. Experimental validation is carried out using a hardware prototype created in the laboratory. The extraction of fundamental currents through the proposed Bell polynomials is compared with least mean square (LMS) and Bernoulli polynomial control algorithms. The NFLC seamlessly carried out grid-to-vehicle (G2V), vehicle–to-grid (V2G) operations in conjunction with a bidirectional DC-DC converter. The dynamic analysis using the BELL polynomial controller is also investigated during load variations.
Diabetic retinopathy (DR) is a leading cause of vision loss in working-age adults. Its diagnosis is particularly challenging due to the asymptomatic nature of the disease in its early stages. Although early detection is essential, current automated diagnostic methods often exhibit limited performance due to subtle early-stage lesions and inter-expert variability, with reported accuracies typically ranging from 85 to 93
Empty Fruit Bunch (EFB), an abundant by-product of the palm oil industry, shows promise as a renewable alternative to conventional cellulose insulation used in oil-immersed transformers. This study evaluates the feasibility of EFB-based insulating paper by comparing its thermal aging behavior with commercial Kraft paper. Accelerated thermal aging was conducted in mineral oil at 90 °C and 140 °C for 30 days. Morphological, physicochemical, mechanical, thermal, and dielectric properties—including degree of polymerization (DP), tensile strength, thermogravimetric analysis (TGA), relative permittivity (RP), alternating current breakdown voltage (ACBDV), and lightning impulse breakdown voltage (LIBDV)—were systematically investigated. Both materials exhibited substantial DP reduction after aging; however, EFB showed slightly lower degradation (19.97
Accurate lung segmentation in Chest X-rays (CXRs) is essential for Computer-Aided Diagnosis systems targeting respiratory diseases. However, current deep learning models present a trade-off: they are either computationally intensive, restricting deployment in resource-constrained clinical settings, or lightweight but compromised in boundary precision. To resolve this challenge, we propose a lightweight hybrid framework addresses the channel-capacity mismatch inherent to MobileNet-based UNet3+ architectures, a structural limitation overlooked in prior lightweight adaptations. Specifically, Squeeze-and-Excitation attention blocks are embedded within full-scale skip connections to recalibrate inter-channel feature responses, enabling lightweight encoders to effectively feed dense multi-scale decoders without an information bottleneck. The framework couples a pre-trained MobileNet-V2 encoder with deep supervision and Test-Time Augmentation to enforce boundary-aware gradient flow and spatial invariance during inference. With 2.55 million parameters, the model achieves a 98.4
Automated analysis of complex visual data remains challenging in settings characterized by high intra-class variability, subtle inter-class differences, and severe class imbalance. In such scenarios, learning from absolute sample representations often leads to unstable and poorly generalizable models. In this work, we propose a change-centric representation learning framework grounded in the principle of intra-inter change interaction learning, which posits that discriminative information is encoded in the behavior of visual differences across intra-class and inter-class sample interactions rather than in isolated appearances. Building on this principle, we introduce the Change Map Siamese Network (CMSN), a metric learning architecture that explicitly models visual divergence between paired samples through learned change maps. The framework is evaluated on the RFMiD2.0 benchmark using a linear probing protocol to isolate the quality of the learned representations. Experimental results demonstrate that the proposed approach achieves strong and stable performance under severe class imbalance, with particularly robust behavior for rare and ultra-rare categories. An ablation study further confirms the importance of explicit change map modeling over conventional feature-based Siamese learning. These findings indicate that intra-inter change interaction learning provides a robust inductive bias for representation learning in imbalanced and low-data regimes.
Electric machine design optimization is challenging due to conflicting objectives and complex, nonlinear relationships among parameters. Conventional optimization methods suffer from poor generalization capability due to limited data availability or incur high computational costs. This study presents a hybrid methodology that integrates Artificial Neural Network and Snake Optimization Algorithm (SO) for the design optimization of low-voltage squirrel cage induction motors. In the first stage, a Recurrent Neural Network is trained on 1164 motor dataset, whose 11 input features are sourced from manufacturer catalogs and whose 28 mechanical output features are derived through analytical computations, spanning a power range from 4 kW to 900 kW to predict the initial design parameters. In the second stage, these predictions serve as initial vectors for the SO Algorithm, which are optimized toward efficiency objectives through the adaptive constraint strategy. This hybrid framework does not require experimental procedures or high computational expenditure. It offers high generalizability, owing to its ability to learn from a diverse dataset. The hybrid pattern, which couples a deep learning predictor with a metaheuristic refinement stage under engineering feasibility constraints, can recur across domains where labeled data are sparse and physical models are available. This methodology combines the rapid prediction capability of machine learning with the global search ability of metaheuristic optimization, enabling fast convergence to feasible optimal design parameters. In scenarios where access to critical design data is limited, this approach provides a reliable starting point for processes such as drive operations and fault diagnosis, thereby contributing to industrial applications.
Immersive 360^∘ video streaming demands accurate viewport prediction under tight latency and bandwidth budgets. A fundamental tension exists between local inference on head-mounted devices (HMDs)—which is fast but inaccurate—and remote inference on edge servers—which is accurate but communication-costly. Existing hierarchical offloading systems either ignore prediction reliability altogether or rely on implicit, delayed reward signals that produce unstable policies and wasteful offloading decisions. The core difficulty is that viewport prediction is a continuous-valued regression task, making standard confidence scores from classification inapplicable, and Bayesian or ensemble-based uncertainty estimates too expensive for HMD deployment. This paper addresses this gap by proposing an uncertainty-aware hierarchical inference framework that exposes a lightweight, closed-form reliability proxy—the variance of recent prediction residuals, computable in 𝒪(H) time—directly to a constrained Markov Decision Process (MDP) controller. This explicit confidence signal guides Deep Q-Network (DQN) (Mnih et al. in Nature 518(7540):529–533, 2015. https://doi.org/10.1038/nature14236 ) and Dueling DQN (Wang et al. In: Proceedings of the 33rd international conference on machine learning (ICML). PMLR, pp 1995–2003, 2016) agents to offload selectively and proactively, without modifying the underlying predictor or introducing probabilistic overhead. Compared with implicit-signal baselines, the proposed approach delivers up to 9.6 γ , and penalty weight λ . The framework is primarily designed for single-step-ahead prediction; multi-step prediction remains an important open challenge discussed in the paper.
Accurate remaining useful life (RUL) prediction of bearings is vital for preventing unexpected failures and enabling predictive maintenance. Exponential degradation models are effective for representing monotonic trends in vibration-derived health indicators (HIs), but conventional approaches assume fixed parameters from historical data, limiting adaptability under varying operating conditions. This paper presents a dynamic empirical Bayes framework that integrates an exponential degradation model with online prior parameter updating. Vibration signals are processed to construct a shifted and log-transformed HI. Initial priors are derived from historical datasets and updated in real time using empirical Bayes estimation as new data becomes available. Conjugate Bayesian inference yields posterior parameter distributions, enabling analytical computation of the RUL probability density function with 95
