This study evaluates data-driven virtual strain sensing in the cylindrical region of a Type IV composite overwrapped pressure vessel (COPV) using the open synchronized record of one vessel and one hydraulic burst test. The analysis is intentionally framed as a single-test comparative case study rather than a general reference standard. Twelve cylinder channels, representing six positions with axial and hoop gauges, were examined through position-holdout reconstruction within the same burst record, pressure-conditioned reconstruction without explicit time, and a within-record high-pressure stress test above 125 bar. Direction-wise same-time, pressure-bin, and isotonic baselines were compared with linear regression, ridge regression, random forest, and multilayer perceptron (MLP) models. The 115,308 long-format rows originate from one synchronized experiment and were therefore treated as correlated observations, not independent experimental replications. Within the cylinder subset, the same-time mean achieved R²=0.9915, while the MLP without coordinates achieved R²=0.9906±0.0002. When explicit time was removed, the pressure-bin baseline remained stronger than the time-free MLP. In the high-pressure stress test, all models deteriorated and the time-free MLP fell to R²=-1.8043. A supplementary dome analysis also showed substantially poorer reconstruction for hoop gauges. These findings indicate that replacement of a missing cylinder gauge is feasible within an already instrumented burst test, but they do not establish cross-vessel or cross-test generalization. Practical deployment will require repeated tests, multiple vessels, mechanics-aware models, and calibrated uncertainty estimates.
This study presents research on wind-induced vibration energy harvesting (WIVEH) systems, focusing on those consisting of a flexible beam with piezoelectric transducers (PZT) coupled with a bluff-body geometries in a low-speed open wind tunnel. Three configurations were examined: a single cylinder, two cylinders connected by a flat wall, and a cylinder-cuboid hybrid. Continuous wavelet transform (CWT) analysis and statistical evaluation of beam deflections were employed to assess vibration characteristics, frequency content, and voltage output across air velocities from 3.5 to 15 m/s. Physical experiments are performed to compare the performance of systems with different configurations of bluff-body shape and to check the agreement of mathematical model results with empirical data. Investigation demonstrates the role of nonlinear aerodynamic excitation mechanisms of the vortex-induced vibrations (VIV), galloping, and fluttering effects. The influence of design parameters and discuss the challenges in modelling complex fluid-structure interactions with piezoelectric. Results show that bluff body geometry strongly influences beam vibration amplitude, frequency content, and energy harvesting efficiency. The two cylinders configuration exhibited the highest vibration amplitudes, with an 89%, and maximum energy output (~140 μW/cm³), increase in variability relative to the single cylinder, but at the cost of excessive oscillations beyond 10 m/s, compromising structural stability. The single-cylinder configuration showed moderate and predictable behavior with exponential voltage growth (~75 μW/cm³), while the cylinder–cuboid hybrid achieved the most robust performance, with 15% lower variability growth compared to the single cylinder and stable operation above 15 m/s. These findings confirm that geometry-driven flow–structure interactions determine the balance between harvested power and operational safety. The results indicate that while two-cylinder configurations may be effective for controlled low-speed applications such as heating, ventilation, and air conditioning (HVAC) ducts, the cylinder-cuboid hybrid is better suited for broadband, variable-speed environments powering internet of things (IoT) devices and wireless sensors.
This study presents a comprehensive comparison of five analytical-numerical algorithms designed to solve boundary value problems involving bodies coated with functionally graded materials (FGM). The investigation focuses on an axisymmetric heat conduction problem about local heating of the surface of a coated half-space. All proposed algorithms leverage the Hankel integral transformation to handle the problem's radial symmetry efficiently. Three of the methods approximate the continuously varying properties of the FGM coating by discretizing it into a finite number of homogeneous or inhomogeneous layers, for which analytical solutions are derived individually. The change in the thermal conductivity coefficient of the inhomogeneous layer along its thickness is described by a linear or exponential function. The fourth approach transforms the original boundary value problem into the Hankel transform domain, where derivatives are approximated using established finite-difference schemes, enabling a numerical solution. The fifth algorithm uses the approximation of the solution in the functionally graded coating using modeling functions, providing an alternative to layer discretization. The performance, accuracy, and computational efficiency of these five algorithms are assessed and discussed to identify their suitability for practical engineering applications involving FGMs.
This study proposes a vision-based line-following robot solution to overcome the limitations of infrared sensors in industrial environments with variable lighting conditions. The system integrates an automatic inverse gamma correction algorithm with Otsu’s thresholding to optimize contrast, ensuring accurate path extraction under diverse light intensities. Leveraging geometric parameters derived from principal component analysis, a Mamdani-type fuzzy logic controller is designed to coordinate movement, maintaining trajectory stability and mitigating mechanical oscillations. A central contribution of this research is the successful implementation of an intelligent control model on a low-cost hardware platform via an off-board processing architecture. Experimental results demonstrate that the system exhibits flexible responsiveness and reliable tracking, effectively overcoming wireless communication latency challenges to ensure operational performance in intralogistics automation tasks.
Malignant melanoma, a highly aggressive form of skin cancer, poses a significant global health challenge due to its rapid progression and high mortality rate if not detected on time. Early diagnosis is crucial for improving patient outcomes. The effectiveness of skin cancer detec-tion still faces serious challenges, like visual inspection that is less accurate and time-consuming. However, deep learning-based models provide early and accurate diagnosis, serving as a supporting tool for dermatologists. Thus, this study focuses on indicating the most suit-able model for skin diseases identification. Three prominent, pre-trained deep learning models, ResNet152, DenseNet201 and EfficientNet-B4, were involved in order to detect benign and malignant melanoma skin lesions. The study was performed utilizing a combined ISIC da-tasets gathered between 2018 and 2020 that consist of dermoscopic images. The above-mentioned deep learning algorithms were verified using accuracy, precision, recall, and F1-score metrics. Moreover, in this study the performance of skin cancer detection was enhanced uti-lizing soft, hard voting, and XGBoost ensemble learning methods. Combining two and three models were verified. The single models ob-tained accuracy at the level of 89.20%, 88.20%, and 90.40% for ResNet152, DenseNet201 and EfficientNet-B4, respectively. The soft vot-ing ensemble, merging ResNet-152 with EfficientNet-B4 or all three models, achieved the highest absolute accuracy of 91.30%, demonstrat-ing superior performance in melanoma diagnosis compared to individual models. Hard voting and XGBoost stated to be less effective in melanoma diagnosis. To confirm that the models were making decisions based on the significant image regions representing skin lesions, a visual explainable technique was applied. Gradient-weighted Class Activation Mapping proved the models to focus their attention to the relevant disease features. These findings underscore the potential of combining individual model strengths through ensemble learning to achieve superior diagnostic performance in melanoma detection, supporting clinicians in making more accurate and timely diagnosis.
